Re: minimac for benh

[Kevin B. Hendricks]

Hi Dean, Ben,
FYI: I just got from a long family vacation and I haven't tried the 
paypal thing yet.  So I still owe you $50 (I will figure out how to use 
it this weekend!).  So if you already have a minimac then please put 
the $50 to whatever else you might need.

Sorry for the delay!
Kevin
On 1-Mar-05, at 11:28 PM, Dean Hamstead wrote:
ben machine
did you get the minimac yet?
i just want to express more hero worship at this point in the hope
that it will motivate you to write more great code so i can continue
to be free of x86's and be super geeky
im sure one day ill be able to release some code that will help
someone, but for now its just cash and adoration :)
Dean
--
WWW: http://dean.bong.com.au  LAN: http://www.bong.com.au
EMAIL: [EMAIL PROTECTED]   or   [EMAIL PROTECTED]
ICQ: 16867613
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact 
[EMAIL PROTECTED]


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: Buy benh (or other) a minimac Re: Hi question about mac - mini

[Kevin B . Hendricks]

Hi,
Put me down for $50 contribution too to get Ben a mini Mac.
Kevin
On Jan 24, 2005, at 8:51 PM, Dean Hamstead wrote:
Actually, ive been meaning to make some sort of contribution to the
fine souls who keep linux-ppc going and had been meaning to contact
benh (or other) about some sort of gift.
on that note, i would be interested in contributing say us$50 towards
a mini-mac for benh (if hed be interested, and if he feels he would
be the best person).
given that they are $499 with free shipping (im sure that a usb  
keyboard,
mouse and a vga monitor shouldnt be hard to dig up), if a few more  
people
are willing to contribute we could get something happening.

ok sure i could donate to debian directly. but it wouldnt be tax
deductable anyway (i think, im not american)
so like, 9 or so more people and bam, ordered.
Dean
On Tue, January 25, 2005 12:44 pm, Dean Hamstead said:
has anyone even put linux on the mac-mini yet?
Dean
On Tue, January 25, 2005 12:42 am, linuxx said:
Anyone know if is posible boot this machine without keyboard,mouse  
and
monitor of course after install linux in it , and how it work under
debian-ppc ?
I would like to get one to have 24h/day on and dont load my too much  
my
laptop ... how can go the hot ?
Thanks .
--
  2.6.10 #4 Sun Jan 23 09:43:49 WET 2005 i686 THANKS FOR YOUR  
TIME
  
_ 
__
/\
\
\_| One Bell System - it works.
|
  |
|
  |
__|_

\_/ 
/
  .
 .|.
  \   |   /
  '.   \  '  /   .'
'. .'```'. .'
<>.:::`...`:::<>
<>:  :<>
<>:  Luis Toro Teijeiro  :<>
<>:  :<>
<>:SysAdmin and Consulting.  :<>
<>:http://jaya.dyndns.org:<>
<>:linuxx at jaya dot dyndns dot org :<>
<>:  :<>
<>:  Windows : Where do you want to go today?.   :<>
<>:  MacOS   : Where do you want to be tomorrow?.:<>
<>:  Linux   : Are you coming, or what?. :<>
<>:  :<>
<>:..:<>
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>

--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact
[EMAIL PROTECTED]


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact
[EMAIL PROTECTED]


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact  
[EMAIL PROTECTED]


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

FYI: OpenOffice.org1.1.2 for PPC Linux now available

[Kevin B . Hendricks]

Hi,

FYI,  OpenOffice.org 1.1.2 for PPC Linux is now available.

The main new features are:
   - lots and lots of bug fixes from OOo 1.1.1
  see 
http://download.openoffice.org/1.1.2/release_notes_1.1.2.html
   - Will work with Blackdown JDK *or* the IBM JDK (just select one 
upon install)

   - Autopilot to easily install MS fonts
   - Autopilot to easily install additional dictionaries
   - Integrated in third party jars to support Flat XML export feautres 
for JDK 1.3.1



The main PPC Linux website is:

http://porting.openoffice.org/linuxppc/ppclinks.html

Many of the mirrors are still just syncing up but it is available on the
YellowDog ftp mirror site.

Additional localized versions will be coming soon.

Please read the README_FIRST to do a proper install.

Hope this helps,

Kevin Hendricks
OpenOffice.org Maintainer for LinuxPPC

Re: ppc64 port

[Kevin B . Hendricks]

Hi,


Let's review where some of the the other distros are at and what they
are doing for x86_64 as well. Everyone (and I mean everyone) has the
ability to run both 32 bit and 64 bit code.


For x86-64 and ia64 this makes sense. There are a good
number of non-free binaries out there -- though I must
admit that Debian traditionally gives non-free binaries
the cold shoulder.

Those of us running ppc have long gotten used to doing
without any of those non-free binaries. There isn't any
compatibility benefit to supporting 32-bit stuff. It's
only a matter of performance: do you want 1 glibc in RAM,



I disagree.

There are many things that may not run or that will run better as 32 
bit.

For example the OpenOffice.org tree is not 64bit clean (or even close)
and will probably take a lot of work to reach that point.  On x86_64 
they

simply use the 386 rpm.

I am sure there are other large peices of code that are not 64 bit 
clean as well.
So it would be nice to run 32 bit Linux PPC OOo version until the tree 
actually gets

cleaned up and the uno-bridge code is written to deal with 64 bit vs
32 bit abi differences.

My 2 cents,

Kevin

Re: Announcing: PPC Linux Version of OpenOffice.org 1.1.1

[Kevin B . Hendricks]

Hi Jason,


If we use the debian packages, can these facilities be added in later?
Or is it too late because they have been compiled out of the code?

I've been trying to use the Report Wizard and the docbook filter and
haven't been able to make them work - this might explain why.


Yes, neither of those filters will work with a non-java build
like the one Debian uses.

AFAIK, official Debian builds do not have the java_uno bridge or
even include it so they will not work with Java. So none of the java
features will work even if you have the Blackdown of IBM java virtual 
machine.


As I said, work is now beginning on trying to use gcj/kaffe so hopefully
with free java versions can help solve this.

Kevin

Re: Announcing: PPC Linux Version of OpenOffice.org 1.1.1

[Kevin B . Hendricks]

Hi Ben,


Is this needed for the build or can it be built and JDK added later ?


The JDK is needed for a full normal build of OOo but Debian and others 
use patches to disable Java since Java itself is not free.  Much of the 
functionality is present without the need for Java.


Actually, my question is simpler: I use debian, I have debian sid 
version
of OpenOffice. I have the IBM JDK in /usr/local, can I setup that OOo 
to

use that JDK or was it built in such a way that no JDK can be used ?



No, I don't think so.  Debian builds disable java completely since it 
is Not Free.
There is work going on to build OOo with gcj/kaffe but this is just 
beginning.
Nut I am sure the Debian-openoffice guys can be much more specific than 
I can.



(BTW. What kind of functionality Java brings to OOo that you don't have
in the first place ?)



Java is currently used both in the build itself and for
- All accessibility features for handicapped and disabled (requires JDK 
1.4.1)

- Flat XML file filter
- All jdbc database connectivity
- Report Wizard features
- Applets and java as used in Html web page editing
- pocket excel and pocket word import features
- XSLTFilter and Validation
- docbook format filter
- the java-uno bridge which allows Java components to talk to
Basic and C++ components via the uno bridge
- Used in the SDK to control and automate OOo
- It is used along with rhino and js to provide scripting for OOo 2.0

There might be a few other things but these are the ones I can think of 
now.
I do think there will be more major java components coming down the 
road in the future.


Kevin

Re: Announcing: PPC Linux Version of OpenOffice.org 1.1.1

[Kevin B . Hendricks]

Hi Ben,

The old JDK is the one I use with OOo.  I know that work is still 
ongoing at Blackdown for PPC Linux since Sun expressed some interest in 
it recently (but not enough to actually find me to do it) so I have 
pretty much given up on it.  Frankly I use the IBM jdk 1.4.1 for almost 
everything.  Unfortunately, the damn OOo source code requires a "Sun" 
style JDK layout.


I recently had to fix OOo so it would work on the MacOSX JDK and now 
that I am familiar with the code I can do the same for IBM's JDK.


They are actually very very close.  In fact, you can actually use 
symlinks to fool OOo to accept the IBM JDK and then go ahead and use 
it.


Once I reboot into PPC Linux I will dig up what symlinks I added and 
send them to you and you can use them temporarily until I get the 
source changes in the OOo tree.


Kevin


On Mar 30, 2004, at 6:09 PM, Benjamin Herrenschmidt wrote:


On Wed, 2004-03-31 at 02:28, Kevin B.Hendricks wrote:

FYI only,

OpenOffice.org announces OpenOffice.org 1.1.1 for PPC Linux is now
available on the StarDiv and YellowDog Linux Mirrors and selected 
other

locations.

http://porting.openoffice.org/linuxppc/ppclinks.html

This contributed build was made on an almost stock YDL 3.0 machine and
therefore should run on any PPC Linux system using glibc 2.3.2.  The
Blackdown JDK is also recommended for full use.


Hi Kevin !

Is there a recent Blackdown JDK for PPC, or are we still stuck at 1.3 ?

Does it have a working JIT nowadays ?

I've been using the IBM one lately for my (few) Java needs...

Ben.

Announcing: PPC Linux Version of OpenOffice.org 1.1.1

[Kevin B . Hendricks]

FYI only,

OpenOffice.org announces OpenOffice.org 1.1.1 for PPC Linux is now 
available on the StarDiv and YellowDog Linux Mirrors and selected other 
locations.


http://porting.openoffice.org/linuxppc/ppclinks.html

This contributed build was made on an almost stock YDL 3.0 machine and 
therefore should run on any PPC Linux system using glibc 2.3.2.  The 
Blackdown JDK is also recommended for full use.


Also available are complete build instructions and source code 
additions for those would would rather build their own (see the 
"source_build_additions" archive).  Localizations for French and many 
other languages will also soon be available.  Check the ppclinks page 
for more info.


Thanks,

Kevin Hendricks
OpenOffice.org Porting Team for PPC Linux and MacOSX

Re: IBM Java SDK 1.4.1

[Kevin B. Hendricks]

Hi,

Possibly?  The Blackdown JDK source does not come with any JIT sources but it 
does come with Hotspot sources but not for PPC.

Would someone like me be considered "contaminated" even if there is no JIT 
being used (or Hotspot for that matter) for PPC since I have seen the built 
the Sun sourcecode as part of Blackdown?  What if I just stick to the ppc 
assembler part?

Kevin


 
On Sunday 25 January 2004 11:51, Dalibor Topic wrote:
> Frank Gevaerts wrote:
> > On Sun, Jan 25, 2004 at 12:12:20AM +0100, Victor Fuente wrote:
> >>Thank you, could you please tell me where is the JIT Documentation??
> >
> > I posted all I know. I found it in the mailing list archives some time
> > ago.
>
> While we are on the subject of JITs for java runtimes for PowerPC
> processors, I'd like to point out that JanosVM 0.8.0[1], a fork of kaffe
> virtual machine [2], contains an alpha version of a JIT for PowerPC[3].
>
> In the hands of a willing PowerPC assembler speaking hacker, it could be
> merged back into the main kaffe sources, and turned into a nice
> perfomance booster for kaffe (which also happens to work fine on power64
> based platforms in the CVS).
>
> Any takers?
>
> cheers,
> dalibor topic
>
> [1] http://freshmeat.net/redir/janosvm/14957/url_tgz/janosvm-0.8.0.tar.gz
> [2] http://www.kaffe.org
> [3] See config/powerpc directory.

Re: bitkeeper trees: where is the debian linux bk for powerpc?

[Kevin B. Hendricks]

Hi,

FWIW,  I just rsyncd and built 2.6.1-rc1-ben1, added a few patches to make my 
aec62XX card work again and edited the pmac_defconfig to enable SMP, IRQ 
sharing, and the aec62XX and rebuilt it and I must say all of my troubles 
with 2.6.0-test11 have gone away.

- once I have loaded the correct modules and set the volume control, sound now 
works

- all of the trouble I had with randomly disconnecting usb devices in ben's 
2.4.23 tree have now gone away (all I had to do was a big build on 2.4.23 and 
walk away and when I came back, no keyboard input was possible without 
disconnecting and reconnecting the keyboard usb cable in 2.4.23)

So all in all, this tree performs better and is more stable then my last 
2.4.23 build from Ben's tree (mainly due to usb disconnect problems in that 
tree whose fixes have not been backported/applied to 2.4.23).

It seems to be a big improvement.  I will probably stay with 2.6.1 and not go 
back to 2.4.23 (at least not until they fix the usb disconnect problems!).

Kevin


On Saturday 03 January 2004 09:59, Jens Schmalzing wrote:
> Hi,
>
> Wolfgang Pfeiffer writes:
> > The mvista one seems out of sync again.  I have rsynced today and
> > got -ben1 where the BitKeeper tree is -ben4.
>
> Both the rsync.theorie.physik.uni-muenchen.de and the
> source.mvista.com mirrors are up to date, at 2.6.1-rc1-ben1.  The
> cause for this confusion is probably a Changeset that first (in
> revision 1.404) bumps the EXTRAVERSION from -ben3 to -ben4, then (in
> revision 1.405) corrects this to -rc1-ben1.
>
> > But is there a powerpc debian-tool for bk?
>
> bk is not free and not redistributable, so there is no Debian package.
>
> And you don't need bk to verify the above anyway, a webbrowser is
> sufficient to compare a few files from the bk tree to what you got
> from the rsync mirrors.
>
> Regards, Jens.
>
> --
> J'qbpbe, le m'en fquz pe j'qbpbe!
> Le veux aimeb et mqubib panz je pézqbpbe je djuz tqtaj!

Re: 2.6.0-ben3: Badness in redraw_screen

[Kevin B . Hendricks]

Hi,

In the first hunk of your patch you can possibly grab the semaphore and 
then do a return 1 with the semaphore held.


If you really need to hold the semaphore to do a vc_allocate then you 
should remember to release that semaphore before doing the return 1


So something along the lines of ...

acquire_console_sem();
if (vc_allocate(SUSPEND_CONSOLE)) {
   release_console_sme();
   return 1;
}

would be better I think if you really do need to hold the console_sem() 
before calling vc_allocate.


Kevin



On Dec 31, 2003, at 9:15 AM, Michael Schmitz wrote:


[CC to debian-powerpc as this was reported there as well ...]


I'm getting this badness in redraw_screen when my iBook wakes up with
2.6.0-ben3:

Badness in redraw_screen at drivers/char/vt.c:596
Call trace:
 [c000bd50] dump_stack+0x18/0x28
 [c0008c44] check_bug_trap+0x84/0xac
 [c0008d34] ProgramCheckException+0xc8/0x180
 [c00082cc] ret_from_except_full+0x0/0x4c
 [c015a94c] redraw_screen+0x2c/0x1e0
 [c00376b4] pm_restore_console+0x38/0x48
 [c039ef14] 0xc039ef14
 [c039f468] 0xc039f468
 [c039fdc8] 0xc039fdc8
 [c00701d0] sys_ioctl+0x278/0x2f4
 [c0007d0c] ret_from_syscall+0x0/0x4c



Similar badness can be had in set_origin (vt.c:568) and set_palette
(vt.c:2851), on my Lombard. That's called from pm_prepare_console, 
for a

change.


(The three unnamed frames are inside pmac_wakeup_devices,
powerbook_sleep_Core99 and pmu_ioctl).  Does pm_restore_console need
to acquire the console lock, or is that the duty of the callers?


None of the other callers does this so I'd guess it needs to be done 
in

pm_restore_console (and pm_prepare_console).


The following patch works for me... should we submit that to lkml?

Michael

--- kernel/power/console.c.org  2003-12-31 14:08:04.0 +0100
+++ kernel/power/console.c  2003-12-31 14:09:26.0 +0100
@@ -20,12 +20,12 @@
 #ifdef SUSPEND_CONSOLE
orig_fgconsole = fg_console;

+   acquire_console_sem();
if (vc_allocate(SUSPEND_CONSOLE))
  /* we can't have a free VC for now. Too bad,
   * we don't want to mess the screen for now. */
return 1;

-   acquire_console_sem();
set_console(SUSPEND_CONSOLE);
release_console_sem();
if (vt_waitactive(SUSPEND_CONSOLE)) {
@@ -42,12 +42,14 @@
 {
console_loglevel = orig_loglevel;
 #ifdef SUSPEND_CONSOLE
+   acquire_console_sem();
set_console(orig_fgconsole);

/* FIXME:
 * This following part is left over from swsusp. Is it really needed?
 */
update_screen(fg_console);
+   release_console_sem();
 #endif
return;
 }


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact 
[EMAIL PROTECTED]

Re: also sound and lost cpu

[Kevin B. Hendricks]

Hi Ben,

It seems top always reports the following on the 2.6-0 test11 kernel

10:30am  up 36 min,  4 users,  load average: 0.85, 0.63, 0.48
77 processes: 74 sleeping, 3 running, 0 zombie, 0 stopped
CPU0 states: 26.0% user,  2.4% system,  0.0% nice, 71.0% idle
fscanf failed on /proc/stat for cpu 1
Mem:   771980K av,  398708K used,  373272K free,   0K shrd,   18908K buff
Swap:


Notice the "fscanf failed on /proc/stat for cpu 1"

I then catted /proc/stat and I can't really see anything wrong

[EMAIL PROTECTED] kbhend]$ cat /proc/stat
cpu  80507 0 9043 354838 7327 99 4
cpu0 43193 0 4267 175145 3232 15 0
cpu1 37313 0 4776 179692 4094 84 4
intr 387114 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 27 8 0 0 0 0 4086 24 73 
287715 0 0 0 0 0 0 0 0 0 0 0 3 677 0 0 0 0 0 0 0 0 0 0 0 14 19955 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 82 74450 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0
ctxt 584779
btime 1070981597
processes 8360
procs_running 3
procs_blocked 0

I can actually add the LC field to top (using f and then selecting the last 
processor used field) and it seems to show both processors humming along 
nicely.

So maybe just some wierd format change in the output used in /proc/stat that 
is causing "top" to be confused?

Kevin

On Monday 08 December 2003 19:38, Kevin B. Hendricks wrote:
> Hi Ben,
>
> I also seem to be having sound troubles with 2.6.0 that do not happen with
> 2.4.23.  I have buyilt 2.6.0-test11 with i2c-keywest built in and not as a
> module so I can't try the rmmod trick.
>
> I will rebuild and revert to keeping it a module.
>
> The real issue is that I have run into one of my 2 cpu's literally
> disappearing according to "top" and to its slower performance.
>
> This has never happened with 2.4.XX.  Has anyone else reported anything
> similar on their SMP machines.
>
> If it matters, this is with HIGHMEM but without distributing irq/interrupts
> to every CPU.
>
> I have since booted back into 2.4.23 and all seems fine again.
>
> Kevin

Re: also sound and lost cpu

[Kevin B. Hendricks]

Hi,

> > The real issue is that I have run into one of my 2 cpu's literally
> > disappearing according to "top" and to its slower performance.
>
> Haven't seen that. Remind me the exact machine model and how
> do you reproduce that ? You can try using xmon-over-firewire
> from another machine to find out where the "other" cpu is
> gone ?

This is a Dual G4 1 gig Hz machine with 768 meg of memory.
As for how to recreate this problem, I am not sure yet...  All I know is that 
when editing a huge 20 meg or so text file in emacs, performance started to 
bog downand so I tried top in another window thinking that some cron job had 
started up and it reported only cpu0 and stated that it could not find/report 
on cpu1.

I ran other tests but could not determine what the problem came from and so I 
eventually restarted and both cpus showed up fine then.  

I will try one more clean rebuild with test11 and this time enable interrupts 
to both processors and try to recreate it in a more controlled manner.

I'll let you know if I find anything recreateable.

Kevin

also sound and lost cpu

[Kevin B. Hendricks]

Hi Ben,

I also seem to be having sound troubles with 2.6.0 that do not happen with 
2.4.23.  I have buyilt 2.6.0-test11 with i2c-keywest built in and not as a 
module so I can't try the rmmod trick.

I will rebuild and revert to keeping it a module.

The real issue is that I have run into one of my 2 cpu's literally 
disappearing according to "top" and to its slower performance.

This has never happened with 2.4.XX.  Has anyone else reported anything 
similar on their SMP machines.

If it matters, this is with HIGHMEM but without distributing irq/interrupts to 
every CPU.

I have since booted back into 2.4.23 and all seems fine again.

Kevin

Re: trouble using IBM 1.4.1 jdk for ppc on Debian

[Kevin B. Hendricks]

Hi,

> Nope, setting JITC_PROCESSOR_TYPE to any value from 0 to 8 doesn't seem
> to help. Setting JAVA_COMPILER=NONE worked, but there is not JIT then.

The code in the JIT looks in /proc/cpuinfo and parses it to try to detect 
processor type.

Is there some strange restriction on your processor type (no floating point 
unit, strange cache line size, problems with sync or isync, cache flushing,or 
...) that might be causing a problem?

Otherwise it should work.

Kevin

Re: trouble using IBM 1.4.1 jdk for ppc on Debian

[Kevin B. Hendricks]

Hi,

If you search the archives here you will find a mail I sent to Ben and a few 
others a while back that describes an environment variable you can set to 
force the JIT to recognize your processor (lie to it and tell it is is a 604e 
system).

I have a dual G4 machine and it now works fine.

I think (but I don't remember anymore) that you should try somehting like:

export JITC_PROCESSOR_TYPE=6

And that may help.

Give it a try and let me know if it does.

Kevin


On Monday 08 December 2003 09:45, Sven Luther wrote:
> On Mon, Dec 08, 2003 at 09:34:54AM -0500, Kevin B. Hendricks wrote:
> > Hi,
> >
> > Make sure you have the very very latest version of the IBM jdk 1.4.1 for
> > p-series.  A previous version had a "bug" that created illegal
> > instruction when using the JIT for some processor types it could not
> > identify.
>
> Ok. This must be it. I have a 750Cxe 600MHz. Also, it is probable that
> the board i use being the Pegasos, it may also cause problem.
>
> > The very latest version of the IBM JDK no longer has that problem on my
> > machine.
>
> Well, i donwloaded it today from the IBM site. I doubt they have a new
> version. I guess i will have to contact their support site or something.
> BTW, what processor are you using ?
>
> Thanks for the information.
>
> Friendly,
>
> Sven Luther

Re: trouble using IBM 1.4.1 jdk for ppc on Debian

[Kevin B. Hendricks]

Hi,

Make sure you have the very very latest version of the IBM jdk 1.4.1 for 
p-series.  A previous version had a "bug" that created illegal instruction 
when using the JIT for some processor types it could not identify.

The very latest version of the IBM JDK no longer has that problem on my 
machine.

Kevin


On Monday 08 December 2003 08:20, Sven Luther wrote:
> On Thu, Nov 13, 2003 at 10:41:43AM +0100, Frank Routier wrote:
> > Sure,
> >
> > you can find it here :
> >
> > https://www6.software.ibm.com/dl/lxdk/lxdk-p
> >
> > To go there, you can also go to www.ibm.com, follow developpers - java
> > technology - IBM jdk
> >
> > Latest 1.4.1 version (SR3 ?) works fine on my Debian Tibook, with for
> > example jonas, ant, jboss, barracuda or netbeans.
>
> Could you tell us more about what you have been doing to make it work.
> Have you used alien and created a .deb out of the .rpm provided by IBM,
> or untarred the tarball somewhere. If so, where did you put it ? In /opt
> as the tarball seemed to want it, or somewhere else ?
>
> For me, when unpacking the JDK .tgz somewhere, setting JAVA_HOME to it,
> and then trying out jboss 3.2.2 for example, it dies with :
>
> $ bin/run.sh
> ===
>=
>
>   JBoss Bootstrap Environment
>
>   JBOSS_HOME: /home/luther/java/jboss-3.2.2
>
>   JAVA: /home/luther/java/IBMJava2-ppc-141/bin/java
>
>   JAVA_OPTS:  -Dprogram.name=run.sh
>
>   CLASSPATH:
> /home/luther/java/jboss-3.2.2/bin/run.jar:/home/luther/java/IBMJava2-ppc-14
>1/lib/tools.jar
>
> ===
>=
>
> JVMDG217: Dump Handler is Processing a Signal - Please Wait.
> JVMDG303: JVM Requesting Java core file
> JVMDG304: Java core file written to
> /home/luther/java/jboss-3.2.2/javacore.20031208.142237.719.txt JVMDG215:
> Dump Handler has Processed Exception Signal 4.
> bin/run.sh: line 192:   714 Instruction illégale   $JAVA $JAVA_OPTS
> -classpath "$JBOSS_CLASSPATH" org.jboss.Main "$@"
>
> So, i wonder what i am doing different from you here, since you are
> clearly telling it works for you.
>
> Friendly,
>
> Sven Luther

Re: upgrade of kernel breaks drm/dri?

[Kevin B. Hendricks]

Hi,

My problem simply went away *after* I remembered to enable the CONFIG_DRM=y in 
the kernel config and rebuilt (and then chose the right driver) the kernel.

DRM is NOT built into the kernel by default.  You must enable it in the kernel 
config first and then rebuild your kernel.

HTH,

Kevin


On Monday 01 December 2003 02:12, Colin Leroy wrote:
> On 30 Nov 2003 at 10h28, Kevin B. Hendricks wrote:
>
> Hi,
>
> > drmOpenDevice: minor is 0
> > drmOpenDevice: node name is /dev/dri/card0
> > drmOpenDevice: open result is -1, (No such device)
> > drmOpenDevice: open result is -1, (No such device)
>
> I have exactly the same behaviour on iBook G4. /dev/dri is empty. I have
> DRI's xfree CVS installed and modules agpgart and radeon are inserted.
> If I try to create the nodes with mknod (c 226 0), startx removes them...
>
> I don't know what to do. Could someone help ? thanks in advance...
> --
> Colin
> for i in `find /home -type f`; do \
>   mv "$i" "`head -c 10 /dev/urandom`"; \
> done;

Re: upgrade of kernel breaks drm/dri?

[Kevin B. Hendricks]

Hi,

Sorry, please ignore this.  I shoudl have double checked that the CONFIG_DRM 
was even abled at all.  I thought it would be default but was not even set.

My bad...

Kevin
 
On Sunday 30 November 2003 10:28, Kevin B. Hendricks wrote:
> Hi,
>
> Forgive me if this has been dicsussed before.  Tux racer used to work just
> fine with my old YDL 3 kernel and XFree86 4.3.0.
>
> I have now upgrade to to Ben's latest 2.4.23 kernel and it seems sadly Tux
> racer has sloowed way down again.
>
> Looking in /var/log/XFree86*.log the only bad thing I can see is:
>
> drmOpenDevice: minor is 0
> drmOpenDevice: node name is /dev/dri/card0
> drmOpenDevice: open result is -1, (No such device)
> drmOpenDevice: open result is -1, (No such device)
> drmOpenDevice: Open failed
> drmOpenDevice: minor is 0
> drmOpenDevice: node name is /dev/dri/card0
> drmOpenDevice: open result is -1, (No such device)
> drmOpenDevice: open result is -1, (No such device)
> drmOpenDevice: Open failed
> [drm] failed to load kernel module "radeon"
> (II) RADEON(0): [drm] drmOpen failed
> (EE) RADEON(0): [dri] DRIScreenInit failed.  Disabling DRI.
>
>
> So it appears I seem to be missing a device node.  I manually used mknod as
> root to create /dev/dri/card0 (card1, card2 and card3 already existed but
> no card0) using char 226,0
>
> But it seems that everytime I startup X something deletes my newly created
> device node "card0".
>
> Does anyone have a clue as to why that special device keeps getting deleted
> and or what I should do to get poor Tux back up to speed.
>
> Any hints appreciated.
>
> Thanks,
>
> Kevin

upgrade of kernel breaks drm/dri?

[Kevin B. Hendricks]

Hi,

Forgive me if this has been dicsussed before.  Tux racer used to work just 
fine with my old YDL 3 kernel and XFree86 4.3.0.

I have now upgrade to to Ben's latest 2.4.23 kernel and it seems sadly Tux 
racer has sloowed way down again.

Looking in /var/log/XFree86*.log the only bad thing I can see is:

drmOpenDevice: minor is 0
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is -1, (No such device)
drmOpenDevice: open result is -1, (No such device)
drmOpenDevice: Open failed
drmOpenDevice: minor is 0
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is -1, (No such device)
drmOpenDevice: open result is -1, (No such device)
drmOpenDevice: Open failed
[drm] failed to load kernel module "radeon"
(II) RADEON(0): [drm] drmOpen failed
(EE) RADEON(0): [dri] DRIScreenInit failed.  Disabling DRI.


So it appears I seem to be missing a device node.  I manually used mknod as 
root to create /dev/dri/card0 (card1, card2 and card3 already existed but no 
card0) using char 226,0

But it seems that everytime I startup X something deletes my newly created 
device node "card0".

Does anyone have a clue as to why that special device keeps getting deleted 
and or what I should do to get poor Tux back up to speed.

Any hints appreciated.

Thanks,

Kevin

Re: jre 1.4 for powerpc

[Kevin B. Hendricks]

Hi,

Yes, I suggest someone port the kde/konqueror way of interfacing to java for 
mozilla.  I will most likely not be making any more Mozilla Plugin or 
Blackdown releases for PPC Linux JDK 1.3.1  We do have people who are working 
on JDK 1.4.1 and hotspot but I am not sure of any timeline.

Frankly Konqueror will nicely interface to the IBM JDK and seems to work well 
without the need for complicated plugins that seem to break periodically by 
the slightest change in toolchain and whatnot.

So perhaps an alternative to the Mozilla plugin (that is closely tied to Sun) 
and that works with Kaffe, the GCJ, and IBM's jdk would be a worthwhile 
project.

My 2 cents,

Kevin
 
On Friday 28 November 2003 11:46, Michael Flaig wrote:
> On Thu, Nov 27, 2003 at 11:51:34PM -0500, Gregory Seidman wrote:
> > IBM provides an RPM of their JRE/JDK for Linux on PPC. It can be alien'd
> > and installed. I don't have a direct link for you, but google is your
> > friend.
>
> AFAIK ibm does not provide an plugin for mozilla (and based) browsers
> with their jdk distribution.
>
> cu,
>
> Michael

Re: CFP (Call for Participation) Debian PPC64

[Kevin B . Hendricks]

Hi,

Speaking of the abi, they did just seem to copy it over blindly from 
IBM arches.


For example ... Linuxppc64 even follow the broken AIX alignment that 
long long ints are aligned to 8 but doubles are only aligned to 4 in 
structures.  This creates lots of problems for OpenOffice.org since AIX 
and Linuxppc64 are the only platforms doing that.  I contributed a 
patch that allows doubles to be aligned to 8 (-malign-natural) to gcc 
that should be part of gcc 3.4 when it is released.  It was accepted.


Perhaps, we should spend some time coming up with some basic  abi  
improvements and try to get patches into the gcc tree to conditionally 
enable them.  I don't think IBM or Apple which uses a variant of IBM's 
abi are going to be helping make any changes on their own.  So I think 
it is up to us if it is going to get done, otherwise we are doomed to 
live with it since it will soon be cast in stone.


My 2 cents,

Kevin

On Nov 16, 2003, at 12:19 AM, Benjamin Herrenschmidt wrote:




It sounds like a reason to get a new ABI. There's
no reason we have to blindly follow OS/400.


Amen !

Ben.

Re: CFP (Call for Participation) Debian PPC64

[Kevin B . Hendricks]

Hi Tom,

I got got a dual Operton system that is similar in many ways to the
PPC64 versus 32 issue in that 32 bit apps run at full speed under 
Opteron.


Using that as an example:

- kernel is 64 bit supporting both 32 but and 64 bit syscalls
- /lib and /lib64 both exist
- tool chain defaults to 64 bit but with addition of -m32 wll generate 
32 bit

- gdb32 and gdb available
- almost all packages are compiled 64 bit src rpms, with just a few key 
libraries done in 32 bit


This system works quite well as a research box, able to compile and run 
32 bit apps like SAS, but also able to compile and use R and my own 
code using -m64 that allow its numberical processing power to shine.


Something like that for G5 from Debian would be very very nice indeed!

One question, why do 32 bit apps run faster than 64 bit apps, is it all 
of the memory accesses needed to load 64 bit immediate and other data 
when restricted to 32 bit instruction lengths?


Just wondering

Kevin

On Nov 11, 2003, at 1:04 PM, [EMAIL PROTECTED] wrote:



On Tuesday, November 11, 2003, at 11:06 AM, Leigh Brown wrote:


Albert Cahalan said:

On Mon, 2003-11-10 at 22:07, [EMAIL PROTECTED] wrote:

  Thankfully this is not a huge undertaking as the goal is not to 
create

an end to end 64 bit system. (Tho that could be done, but perhaps
that's a discussion for another day)


It's not as if the Linux apps are all 32-bit.
Due to the Alpha, Opteron, and Itanium, the
code should all be clean.

You'll be running a mix if you go this route,
keeping both libraries in memory. Yuck. Some
apps need to be 64-bit. For example, procps
must be compiled as a 64-bit executable if
you want to support a 64-bit kernel.


Not true, at least on ppc64.  I've run debian on my power3
machine with a 64-bit kernel, and procps works fine.  In fact,
everything runs fine.  This begs the question, what will a ppc64
Debian look like?  I'd guess a 64-bit kernel, 32-bit and 64-bit
toolchains,  ?


Hmm I checked my Power3 box with SuSE on it and in that case they
had it built 64 bit. Still no biggie, I think you're right Leigh,
shouldn't be alot needed that's 64 bit ... least not initially. But 
certainly

in time that can (and should) change.


I hope you'll agree this is a worthwhile endeavor and with your
support, debian support for PowerPC64 processors can be enjoyed by 
all.

(Or at least those with or soon to have hardware ;-)


Eh, does Linux even run on Apple's new box?


Apparently so, but pSeries boxes should also be supported.


I agree, and don't forget iSeries boxes.

Basically I'd like to see supported:

Native on the metal, Apple G5, Power3, RS64-III, RS64-IV, Power4
and for those boxes that support it, Logical Partition iSeries & 
pSeries


Regards,

Tom


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact 
[EMAIL PROTECTED]

Re: java with mozilla-firebird & mozilla

[Kevin B. Hendricks]

Hi,

It can't seem to find the mozilla libs (libxpcom.so and libnspr4.so).
Do you know why not?

I just tried using my Mozilla 1.3 build (admittedly old) with the Blackdown 
JDK 1.3.1 plugin on the URL

http://www.barbery.net/anagram/

and it starts up fine (although I can't read that language)

I was also able to use Konqueror with the Blackdown JDK 1.3.1 and access the 
same site with no problems.

So whatever problem you are having seems to be specific to your machine.

Kevin


On Monday 10 November 2003 14:16, Vincent Lefevre wrote:
> On 2003-11-10 13:12:02 -0500, Kevin.Hendricks wrote:
> > Let's see what the javaplugin is really dynamically linked with on
> > your machine
> >
> > cd j2sdk1.3.1/jre/plugin/ppc/mozilla/
> > ldd javaplugin_oji.so
>
> ay:~> cd /usr/local/j2re1.3.1/plugin/ppc/mozilla
> ay:...lugin/ppc/mozilla> ldd javaplugin_oji.so
> libxpcom.so => not found
> libnspr4.so => not found
> libgtk-1.2.so.0 => /usr/lib/libgtk-1.2.so.0 (0x6fde8000)
> libgdk-1.2.so.0 => /usr/lib/libgdk-1.2.so.0 (0x6fd8b000)
> libgmodule-1.2.so.0 => /usr/lib/libgmodule-1.2.so.0 (0x6fd67000)
> libglib-1.2.so.0 => /usr/lib/libglib-1.2.so.0 (0x6fd1b000)
> libdl.so.2 => /lib/libdl.so.2 (0x6fcf8000)
> libXi.so.6 => /usr/X11R6/lib/libXi.so.6 (0x6fccf000)
> libXext.so.6 => /usr/X11R6/lib/libXext.so.6 (0x6fc9d000)
> libX11.so.6 => /usr/X11R6/lib/libX11.so.6 (0x6fba5000)
> libstdc++.so.5 => /usr/lib/libstdc++.so.5 (0x6fabe000)
> libm.so.6 => /lib/libm.so.6 (0x6fa29000)
> libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x6f9fc000)
> libc.so.6 => /lib/libc.so.6 (0x6f8a3000)
> /lib/ld.so.1 => /lib/ld.so.1 (0x0800)
>
> --
> Vincent Lefèvre <[EMAIL PROTECTED]> - Web:  - 100%
> validated (X)HTML - Acorn Risc PC, Yellow Pig 17, Championnat International
> des Jeux Mathématiques et Logiques, TETRHEX, etc.
> Work: CR INRIA - computer arithmetic / SPACES project at LORIA

Re: java with mozilla-firebird & mozilla

[Kevin B. Hendricks]

Hi,

Not if it is the one I put on the YellowDog Linux mirrors in OpenOffice.org.

That was compiled with gcc 3.2.2 and does work with mozilla compiled with gcc 
3.2.2 as well as OOo compiled with gcc 3.2.2.

That is why I put it there.

Kevin

On Monday 10 November 2003 04:28, Vincent Lefevre wrote:
> On 2003-11-08 19:14:26 -0500, Rob Latham wrote:
> > No debs, but check out the 'contrib/linuxppc' directory on any
> > OpenOffice.org office mirror.
>
> This is the same version as the one found on Blackdown's site.
> It has been compiled with an old gcc version and doesn't work
> with Mozilla compiled with the current gcc version.
>
> --
> Vincent Lefèvre <[EMAIL PROTECTED]> - Web:  - 100%
> validated (X)HTML - Acorn Risc PC, Yellow Pig 17, Championnat International
> des Jeux Mathématiques et Logiques, TETRHEX, etc.
> Work: CR INRIA - computer arithmetic / SPACES project at LORIA

Re: openoffice.org in Sid

[Kevin B. Hendricks]

FYI:

Tomorrow  OOo 1.1.0 final for PPC Linux will/should be available on the 
YellowDogLinux mirrors in pub/yellowdog/software/openoffice/

These are NOT debs but instead the Official OOo installer build done for PPC 
Linux.

It should work on any system using late versions of glibc-2.3.1 (such as YDL 
3.0) or glibc-2.3.2 such as Mandrake, etc.

You may want to install and use it while you wait for official debian debs to 
come along.

Hope this helps,

Kevin

On Wednesday 01 October 2003 18:35, Martin-Éric Racine wrote:
> On Wed, 1 Oct 2003, Frank Murphy wrote:
> > On Wednesday 01 October 2003 9:36, Martin-Éric Racine wrote:
> > > As it so happens, the official 1.1.0 version is finally released, so
> > > I'm guessing that we can expect packages pretty soon, at this point. :)
> >
> > Actually, there's a build problem with powerpc. See bug 206238 for more
> > info.
> >
> > For the time being, grab the OOo 1.0.2 debs from testing.
>
> Been there, done that, have been hating it for months. And that would be
> 1.0.3.
>
> --
> Martin-Éric Racine
> http://www.pp.fishpool.fi/~q-funk/

Re: Enabling IBM Java 141 for web browsers

[Kevin B. Hendricks]

Hi,

As far as I know, the IBM jdks do not come with a Mozilla plugin so they won't 
work with Mozilla period.

You should be able to use in in Konqueror just by opening it up and editing 
its setting and changing the path to find java in the settings.

>From there you should be good to go.

If you want to use the JIT compiler under IBM's jdk please put the following 
in your .bashrc or .profile

export JITC_PROCESSOR_TYPE=6

Hope this helps,

Kevin

On Tuesday 16 September 2003 14:19, Pander wrote:
> Hi all,
>
> I have installed IBMJava2-ppc-141 and it works great although it doens't
> run applets in mozilla and konquerer. What do I need to do to get this
> working?
>
> I have already
> JAVA_COMPILER=NONE
> JAVA_HOME=/opt/IBMJava2-ppc-141
>
>
> Thanks,
>
> Pander

Re: [Q]IBM's Java 1.4.1 with IBM's JIT

[Kevin B. Hendricks]

Hi,
Just check out the YDL mirrors in software/openoffice/ the last rebuild of the 
JDK is there.

Kevin

On Tuesday 02 September 2003 09:33, Vincent Bernat wrote:
> OoO En ce début d'après-midi nuageux du mardi 02 septembre 2003, vers
>
> 14:27, "Kevin B. Hendricks" <[EMAIL PROTECTED]> disait:
> > Of course there is the Balckdown JDK 1.3.1 for PPC Linux that does work
> > (I use it to build OOo with)
>
> Where did you get it ? The one I have is compiled with gcc 2.95, is
> there a 3.2 version ?
> --
> Don't just echo the code with comments - make every comment count.
> - The Elements of Programming Style (Kernighan & Plaugher)

Re: [Q]IBM's Java 1.4.1 with IBM's JIT

[Kevin B. Hendricks]

Hi Sven,

> > Of course there is the Balckdown JDK 1.3.1 for PPC Linux that does work
> > (I use it to build OOo with) but you are right the Blackdown organization
> > really
>
> You are speaking about OOo 1.0.3 or the new 1.1 stuff ?

I maintain the Linux PPC port for OOo (...not for Debian, Jan does this for 
Debian I do it for upstream).   So I have used the Blackdown JDK 1.3.1 to 
help build all of the OOo releases since the early 609 builds (pre OOo 1.0) 
all the way to the 645 builds (OOo 1.1 RC4 on my machine right now hopefully 
about to be released to become OOo 1.1 final).

Are you interested in making your own OOo builds?

Kevin

Re: [Q]IBM's Java 1.4.1 with IBM's JIT

[Kevin B. Hendricks]

Hi,

An aside ...  we (ppc Linux) are really too small of a development community 
and we should really have just one shared development list to get what few of 
us are still active at least seeing/hearing what the others are doing.

Kevin

On Tuesday 02 September 2003 08:16, Kevin B. Hendricks wrote:
> Hi,
>
> You might want to check out the dev list at YellowDogLinux.
>
> it seems by IBM decided to NOT allow its JIT to run with G4 and other
> non-IBM processors.
>
> There was a bit of a storm (i.e. I complained loudly and so did others on
> that list and on the IBM java newsgroup).  I then had to recompile my
> kernel to report I was using a 604e in /proc/cpuinfo and I was able to
> trick JIT into working.
>
> They said it was all a "mistake" and would fix it in the next release
> (AFAIK - that has not come yet).
>
> Later on someone pointed out an environement variable you can set to
> override the processor detection code in the JIT.
>
> ---snip---
> There is a workaround, which is to set the following environment
> variable:
>
> export JITC_PROCESSOR_TYPE= 6
>
> which will force the 604 class cpu code to be employed.
> ---snip---
>
> If you are interested in the whole sordid tale check out:
>
> The thread starts here:
> http://lists.terrasoftsolutions.com/pipermail/yellowdog-devel/2003-June/000
>209.html
>
>
> The environment variable workaround is here:
> http://lists.terrasoftsolutions.com/pipermail/yellowdog-devel/2003-July/000
>251.html
>
> Hope this helps,
>
> Kevin
>
> On Tuesday 02 September 2003 05:41, Benjamin Herrenschmidt wrote:
> > On Tue, 2003-09-02 at 04:44, W. Crowshaw wrote:
> > > Hi,
> > >
> > > Has anybody managed to get IBM's Java 1.4.1 (sadly,
> > > the only one available to us ppc'ers) working with
> > > the JIT its supplies.  Its readme docs say that all
> > > you need to do is set the JAVA_COMPILER enviromental
> > > variable to: jitc.  To test the docs say simply type:
> > > java -version and it will say "(JIT enabled: jitc)"
> > > Unfortunately, this doesn't work.  I have to have
> > > the gij-3.0 package installed instead and the
> > > JAVA_COMPILER=NONE set for java to run at all.
> >
> > I'm getting JVM crashes when trying to use the JIT too
> >
> > Ben.

Re: [Q]IBM's Java 1.4.1 with IBM's JIT

[Kevin B. Hendricks]

Hi,

> > > (sadly, the only one available to us ppc'ers)

Of course there is the Balckdown JDK 1.3.1 for PPC Linux that does work (I use 
it to build OOo with) but you are right the Blackdown organization really 
could use some pleople that understand the PPC instruction set well-enough to 
help out on the HotSpot port (and yes there are 2 people working on it now 
part-time but progress is really slow).

So if you would like to see an alternative build, please volunteer to help 
write it.

BTW: IBM research jvm will run on PPC Linux, as does Kaffe, and as does 
gcj/gij and you can even remove the SWT code form Eclipse and compile it with 
gcj and get graphical widget code compiled to all native code actually 
working.

Kevin

Re: [Q]IBM's Java 1.4.1 with IBM's JIT

[Kevin B. Hendricks]

Hi,

You might want to check out the dev list at YellowDogLinux.

it seems by IBM decided to NOT allow its JIT to run with G4 and other non-IBM 
processors.

There was a bit of a storm (i.e. I complained loudly and so did others on that 
list and on the IBM java newsgroup).  I then had to recompile my kernel to 
report I was using a 604e in /proc/cpuinfo and I was able to trick JIT into 
working.  

They said it was all a "mistake" and would fix it in the next release (AFAIK - 
that has not come yet).

Later on someone pointed out an environement variable you can set to override 
the processor detection code in the JIT.

---snip---
There is a workaround, which is to set the following environment
variable:

export JITC_PROCESSOR_TYPE= 6

which will force the 604 class cpu code to be employed.
---snip---

If you are interested in the whole sordid tale check out:

The thread starts here:
http://lists.terrasoftsolutions.com/pipermail/yellowdog-devel/2003-June/000209.html


The environment variable workaround is here:
http://lists.terrasoftsolutions.com/pipermail/yellowdog-devel/2003-July/000251.html

Hope this helps,

Kevin

On Tuesday 02 September 2003 05:41, Benjamin Herrenschmidt wrote:
> On Tue, 2003-09-02 at 04:44, W. Crowshaw wrote:
> > Hi,
> >
> > Has anybody managed to get IBM's Java 1.4.1 (sadly,
> > the only one available to us ppc'ers) working with
> > the JIT its supplies.  Its readme docs say that all
> > you need to do is set the JAVA_COMPILER enviromental
> > variable to: jitc.  To test the docs say simply type:
> > java -version and it will say "(JIT enabled: jitc)"
> > Unfortunately, this doesn't work.  I have to have
> > the gij-3.0 package installed instead and the
> > JAVA_COMPILER=NONE set for java to run at all.
>
> I'm getting JVM crashes when trying to use the JIT too
>
> Ben.

Re: 2.4.21-ben2 and acard aec6280m (CONFIG_AEC62XX_TUNING)

[Kevin B. Hendricks]

Hi,

I was looking at 2.4.21-ben1 and there is a old driver in 
drivers/ide/aec62xx.c and one in drivers/ide/pci/aec62xx.[hc]

So I was looking at the old driver source and not the new.  

I have both drivers and will try to figure out what changed and is messing us 
up and send it to the list.

Kevin
  
On Sunday 06 July 2003 15:23, Benjamin Herrenschmidt wrote:
> On Sat, 2003-07-05 at 18:22, Kevin B. Hendricks wrote:
> > Hi,
> >
> > Yes the CONFIG_AEC62XX_TUNING *must* be set for that driver to work for
> > ppc. Without tuning it will not properly work for ppc linux and you get
> > either partition check problems or:
>
> I wonder what driver you are using...
>
> There is no longer drivers/ide/aec62xx.c in 2.4.21* but rather
> drivers/ide/pci/aec62xx.c, which doesn't have a CONFIG_*_TUNING but
> will always try to tune.
>
> Ben.

Re: 2.4.21-ben2 and acard aec6280m (CONFIG_AEC62XX_TUNING)

[Kevin B. Hendricks]

Hi,

Whoops my tree is 2.4.21-ben1 and I just looked in drivers/ide/pci and sure 
enough there are actually two aec62xx drivers in that tree (one there and one 
in drivers/ide).

So the update to the new driver in ide/pci seems to have broken this driver 
for us.

Will you please file a breakage report with Andre (the maintainer) and point 
out his latest changes broke the driver for us.  We really need to get it 
fixed for both this tree and the newer kernel trees as well.
 
I will diff the old and new driver to see if I can figure out what 
changed/broke and try to provide a patch for the new driver to fix it since I 
have a 6280m card to play/test with.

Thanks,

Kevin

On Saturday 05 July 2003 23:45, dorian wrote:
> hmmm...
> as far as i can tell, the CONFIG_AEC62XX_TUNING option is no longer
> there in 2.4.21. the file aec62XX.c has moved (to drivers/ide/pci/) and
> is now at version 0.11 (compared to 0.09 in 2.4.21).
>
> well, i reverted back to 2.4.20-ben10 and everything works fine, so
> it's not really an issue...
>
> thanks anyway,
>
> -dorian

Re: 2.4.21-ben2 and acard aec6280m (CONFIG_AEC62XX_TUNING)

[Kevin B. Hendricks]

Hi,

Yes the CONFIG_AEC62XX_TUNING *must* be set for that driver to work for ppc.  
Without tuning it will not properly work for ppc linux and you get either 
partition check problems or:

> hdX: dma_timer_expiry: dma status == 0x24
> hdX: lost interrupt
> hdX: dma_intr: bad DMA status (dma_stat=30)
> hdX: dma_intr: status=0x50 { DriveReady SeekComplete }


Ben can you please add back in CONFIG_AEC62XX_TUNING and make it the default 
under ppc linux.

Luckily the code in the kernel  in drivers/ide/aec62xx.c
still has the ifdefs in place 

#ifdef CONFIG_AEC62XX_TUNING

struct chipset_bus_clock_list_entry {
bytexfer_speed;

bytechipset_settings_34;
byteultra_settings_34;

bytechipset_settings_33;
byteultra_settings_33;
};


So you could try adding CONFIG_AEC62XX_TUNING=y  to the .config for the kernel 
build and if that doesn't work simply patch drivers/ide/aec62xx.c 

To change every occurence of 

#ifdef CONFIG_AEC62XX_TUNING

to 

#if 1 /* def CONFIG_AEC62XX_TUNING */

And build it yourself.

Kevin

Re: Supported systems (Was: Apple PowerMac G5)

[Kevin B. Hendricks]

Hi,

> Well, they do... internally to Apple ;) Apple is quite clear in it's
> total absence of interest about Linux on their machines. Some ppl here
> even consider us as a competitor...

And I remember all of the time and effort many of us in the community put in 
to help Apple in its MkLinux project when they were using a pre glibc-2 build 
and had to hand apply kernel patches to integrate the Mach kernel and the 
linux kernel/client, etc. 

I remember reminding Apple about that when I asked if they would share their 
HotSpot port with Blackdown (both signed Sun's JDK agreement so there were no 
license concerns) and basically got shit-on in response.

Sometimes I think that Apple (and most other companies) do not understand the 
impact of community goodwill, and its true value.  And Apple is not alone - 
IBM's JDK 1.4.1 (due to a "programmers mistake" ;-)) actually checks 
/proc/cpuinfo under Linux PPC and decided it would abort if a G4 was found
Although they have promised to fix it, the version I downloaded today is still 
the same "broken" one.   I seemed to be one of the few that was outraged by 
that behavior.

Waiting on big-name companies to "help out" is a big mistake.  We are on our 
own and we need to become even more self-sufficient IMHO.

Kevin


  

Re: some basic problems with ppc assembler

[Kevin B. Hendricks]

Hi,

Grab http://crunch.ivey.uwo.ca/khendricks/PPC_docs.tar.gz
There is an old  ABI document.

Please note.  PPC 32 does not follow the published ABI all of the time:

1. does not return small structures (size of 8 or less) in registers

2. does not pass floating point signals as doubles aligned to 8 on the 
outgoing parameter stack.

Have fun!

Kevin

On Wednesday 25 June 2003 16:56, Brad Boyer wrote:
> On Wed, Jun 25, 2003 at 05:02:40PM +, paubert wrote:
> > It is specified in the ABI for functions with variable number of
> > argumentsi: the crxor indicates that no parameters have been passed in
> > the floating point registers so an integer only task will not save
> > fpr1-fpr8 to walk the argument list in va_arg. OTOH if you call a
> > variadic functions with floating point parameters, the call will be
> > preceded by creqv 6,6,6.
>
> Do you actually have a document for the official ABI? I found a link
> to a document in some old messages, but it was a dead link. A generic
> search of the web didn't turn up anything, although I might have
> been searching for the wrong words...
>
> I've been trying to piece it together based on gcc, but that's
> kind of a clumsy way to figure it out.
>
>   Brad Boyer
>   [EMAIL PROTECTED]

Re: 2.4.21 on a G3

[Kevin B. Hendricks]

Hi,

You don't need to do that at all.  Simply rebuild a kernel with both of these 
switches enabled and your 6280M should happily be recognized and work:

CONFIG_BLK_DEV_AEC62XX=y
CONFIG_AEC62XX_TUNING=y

You need both (without the TUNING set it hangs on the aprtition check).
This has worked fine on my 6280M card.

Kevin

On Tuesday 17 June 2003 10:05, Jeroen Roovers wrote:
> On 17 Jun 2003 at 15:22, Jens Francke wrote:
> > hi,
> >
> > i am using AEC6280M in a beige G3.
> > with 2.4.20 i needed benH´s tree to make it work.
> >
> > i just compiled the 2.4.21 vanilla last weekend. AEC6280M works like a
> > charme without any patches:)
>
> With no funny settings?
>
> > give it a try
>
> I will, but the procedure is still quite difficult: the installation
> kernel doesn't like the AEC6280M, as it simply stops even before it
> mounts the root partition, even if no devices are attached to the
> AEC6280M. That means I have to install on a plain system, build a
> kernel that knows about AEC6280M (and doesn't think it's a AEC6280R
> by mistake) then subtly introduce the AEC6280M and maybe quietly
> attach an hd after that. When all that is done, I may be able to copy
> the system to the true destination on the (large) hd attached to the
> AEC6280M, and hopefully still have a running system.
>
> If it isn't openfirmware that gets lost in the process, it's Debian
> PPC (vanilla woody, at 64kbits/s no bandwidth here to update it
> properly) that throws a fit. When I get something of a system going,
> I'll definitely give 2.4.21 a good look.
>
>
>
>Jeroen

Re: Acard AEC6280 / ATP-865

[Kevin B. Hendricks]

Hi,

I uses the ACard 133 ATA pci card on my machine with no problems
as long as you compile your kernel with both:

CONFIG_BLK_DEV_AEC62XX=y
CONFIG_AEC62XX_TUNING=y

That last one is crucial.  Without TUNING=y my box will always hang on the 
partition check for that card.

Here is what dmesg shows ...

AEC6280R: IDE controller on PCI bus 10 dev a0
AEC6280R: chipset revision 2
AEC6280R: not 100% native mode: will probe irqs later
AEC6280R: ROM enabled at 0x8009
ide2: BM-DMA at 0x0400-0x0407, BIOS settings: hde:pio, hdf:pio
ide3: BM-DMA at 0x0408-0x040f, BIOS settings: hdg:pio, hdh:pio

In the OF device tree it looks like the following:
cd /proc/device-tree/@f200/ACARD,[EMAIL PROTECTED]

cd ACARD,[EMAIL PROTECTED]

hexdump device-id
000  0009

hexdump vendor-id
000  1191

So I would say you and I have exact same card.

BTW: If you use this card to hold your root partition then make sure you have 
built your own installer kernel that has both config settings in place 
otherwise it will not boot (I ran into this when I upgraded to YDL 3.0 but 
their kernel does not include the TUNING config option and I could therefore 
not get past the partition check.

With no extra options, after tuning it runs at UDMA(4).  I have not tried 
UDMA(5) since my drive is not rated at that speed.

Hope this helps,

Kevin

On Monday 26 May 2003 02:44, Jeroen Roovers wrote:
> I've seen many questions posted on public message boards and mailing
> list archives about Linux support for the Acard AEC6280 PCI card /
> ATP-865 Ultra ATA 133 controller chip, possibly marked:
>
> Vendor: 0x1191
> Device ID: 0x0009
>
> but I haven't seen support for it expressed anywhere, particularly
> not in response to those public questions. I ran a Debian current
> (3.0r1) boot cd on a Macintosh B/W G3 last night, and neither of the
> installation kernels finds the controller or the disk attached to it.
> If I could find something like an acard.o somewhere, I could probably
> put it on an ext2 partition on an ATA Flash disk, boot from the first
> Debian cd-rom and use insmod to have the controller and disk
> recognised, and then mount a couple of Linux partitions on the hdd to
> get the Debian installation going, but for now, there seems to be no
> support at all from the 2.4.x kernel in Debian 3.0r1.
>
> Was this issue ever resolved?
>
>
>Jeroen

Re: Blackdown JDK and sid (was: Re: java 1.4 for ppc)

[Kevin B. Hendricks]

Hi,

j2sdk1.3.1-02d is compiled with gcc 3.2.2

Please try it with a mozilla compiled with gcc 3.2.2 and
and with gcc-2.3.1 with the proper compatibility patch for the compiler it 
was built with.

This does work (it works fine on YDL 3.0).

On April 21, 2003 07:27 pm, Wes Morgan wrote:
> I installed j2sdk1.3.1-02d but the behavior didn't change any. The plugin 
> still only worked with my 2.95-compiled copy of Mozilla, and I still got 
the 
> relocation error. gcc -v gives "gcc version 3.2.3 20030407 (Debian 
> prerelease)" but I have no idea what version was used to compile the 
Debian 
> packages.
> 
> Wes Morgan
> 
> On Monday 21 April 2003 4:46 pm, Kevin B. Hendricks wrote:
> > Hi,
> >
> > Two things:
> >
> > 1. Mozilla is C++ and as such Mozilla and Blackdown JDK (has C+ in it 
also)
> > must both either be compiled with gcc 2.95.X or both compiled with gcc 
3.X
> > since the C++ ABI changed.
> >
> > 2. The current JDK *does* work with a properly compiled glibc-2.3.1
> > compiled with the same compiler.  For example Blackdown JDK works fine 
with
> > YDL 3.0 which uses glibc-2.3.1 and everything was compiled with gcc 
3.2.X.
> >
> > So the only solution I can offer:
> >
> > - upgrade to the very latest gcc 3.2.X
> > - rebuild glibc 2.3.1 with it
> > - rebuild Mozilla with it
> >
> > Then grab the Blackdown JDK wich was built with the same compiler
> > gcc 3.2.X from
> >
> > http://crunch.ivey.uwo.ca/khendricks/openoffice/
> >
> > Then you should have a working Blackdown JDK, a working Mozilla and you
> > should not be seeing any __dividi3 relocation errors if the proper 
glibc
> > compatibility patches are in place.
> >
> > I realize all of this is a mess but the number of changes in compilers, 
c++
> > abis, and glibc versions (and it matters which gcc cmpiles which glibc 
due
> > to an earlier issue of some symbols sneaking in to the glibc builds 
from
> > gcc) have made it very hard to keep a working Blackdown JDK (especially
> > one that works with a recent Mozilla).
> >
> > Hope that helps,
> >
> > Kevin
> >
> > On April 21, 2003 03:59 pm, Wes Morgan wrote:
> > > Is there any way to get the Blackdown JDK for PowerPC to work with 
Debian
> >
> > sid?
> >
> > > It seems to be a combination of a couple things. Prior to 02c, it 
seemed
> >
> > to
> >
> > > be a gcc-2.95 vs. gcc-3.2 issue, but now it seems to be a libc-2.3 
issue
> >
> > (but
> >
> > > the compiler issue might still be there). One error in particular I 
keep
> > > running into is this:
> > >
> > > /usr/local/j2sdk1.3.1/bin/ppc/native_threads/jar: relocation error:
> > > /usr/local/j2sdk1.3.1/jre/lib/ppc/classic/libjvm.so: symbol __divdi3,
> >
> > version
> >
> > > GLIBC_2.0 not defined in file libc.so.6 with link time reference
> > >
> > > libc.so.6 is pointing to libc-2.3.1.so in /lib. I also cannot get the
> >
> > plugin
> >
> > > to work with current versions of the Mozilla package. If I download
> >
> > Mozilla
> >
> > > 1.3 compiled with gcc-2.95, then the plugin works, so that makes me 
think
> > > that there are still compiler version issues with 02c. But I could be
> >
> > totally
> >
> > > off on all of this... :)
> > >
> > > Wes Morgan
> > >
> > > On Monday 21 April 2003 8:04 am, Kevin B. Hendricks wrote:
> > > > Hi,
> > > >
> > > > AFAIK, none exists.  I am basically the Blackdown JDK for ppc 
porting
> >
> > team
> >
> > > > and I simply have not had the time to start working on JDK 1.4 yet.  
My
> > > > real need is for help from someone who understands PPC assmebler 
and
> > > > instruction set well and can help implement HotSpot for 1.4.
> > > >
> > > > IBM has a wonderful JDK with a fast JIT that I have used but it 
seems
> >
> > to be
> >
> > > > a JDK 1.3.1 version only.  As far as I know they have not even 
released
> >
> > an
> >
> > > > alpha or beta of JDK 1.4 for ppc linux.
> > > >
> > > > Sorry I can't be more help.
> > > >
> > > > Kevin
> > > >
> > > > On April 21, 2003 07:18 am, florian wrote:
> > > > > hi!
> > > > >
> > > > > does anybody know of a java 1.4 implementation for
> > > > > ppc? i searched around quite a bit. but didnt find anything..
> > > > >
> > > > > anybody knows more?
> > > > >
> > > > > ciao!
> > > > > florian
> > > > >
> > > > >
> > > > > --
> > > > > To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> > > > > with a subject of "unsubscribe". Trouble? Contact
> > > >
> > > > [EMAIL PROTECTED]
> > >
> > > --
> > > The most subversive thing you can do is assign full humanity to each 
and
> >
> > every
> >
> > > individual on this planet.
> 
> -- 
> The most subversive thing you can do is assign full humanity to each and 
every 
> individual on this planet.
> 
> 
> -- 
> To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> with a subject of "unsubscribe". Trouble? Contact 
[EMAIL PROTECTED]
> 
> 

Re: Blackdown JDK and sid (was: Re: java 1.4 for ppc)

[Kevin B. Hendricks]

Hi,

Two things:

1. Mozilla is C++ and as such Mozilla and Blackdown JDK (has C+ in it also) 
must both either be compiled with gcc 2.95.X or both compiled with gcc 3.X
since the C++ ABI changed.

2. The current JDK *does* work with a properly compiled glibc-2.3.1
compiled with the same compiler.  For example Blackdown JDK works fine with 
YDL 3.0 which uses glibc-2.3.1 and everything was compiled with gcc 3.2.X. 

So the only solution I can offer:

- upgrade to the very latest gcc 3.2.X
- rebuild glibc 2.3.1 with it
- rebuild Mozilla with it

Then grab the Blackdown JDK wich was built with the same compiler
gcc 3.2.X from

http://crunch.ivey.uwo.ca/khendricks/openoffice/

Then you should have a working Blackdown JDK, a working Mozilla and you 
should not be seeing any __dividi3 relocation errors if the proper glibc 
compatibility patches are in place.

I realize all of this is a mess but the number of changes in compilers, c++ 
abis, and glibc versions (and it matters which gcc cmpiles which glibc due 
to an earlier issue of some symbols sneaking in to the glibc builds from 
gcc) have made it very hard to keep a working Blackdown JDK (especially 
one that works with a recent Mozilla).

Hope that helps,

Kevin

On April 21, 2003 03:59 pm, Wes Morgan wrote:
> Is there any way to get the Blackdown JDK for PowerPC to work with Debian 
sid? 
> It seems to be a combination of a couple things. Prior to 02c, it seemed 
to 
> be a gcc-2.95 vs. gcc-3.2 issue, but now it seems to be a libc-2.3 issue 
(but 
> the compiler issue might still be there). One error in particular I keep 
> running into is this:
> 
> /usr/local/j2sdk1.3.1/bin/ppc/native_threads/jar: relocation error: 
> /usr/local/j2sdk1.3.1/jre/lib/ppc/classic/libjvm.so: symbol __divdi3, 
version 
> GLIBC_2.0 not defined in file libc.so.6 with link time reference
> 
> libc.so.6 is pointing to libc-2.3.1.so in /lib. I also cannot get the 
plugin 
> to work with current versions of the Mozilla package. If I download 
Mozilla 
> 1.3 compiled with gcc-2.95, then the plugin works, so that makes me think 
> that there are still compiler version issues with 02c. But I could be 
totally 
> off on all of this... :)
> 
> Wes Morgan
> 
> On Monday 21 April 2003 8:04 am, Kevin B. Hendricks wrote:
> > Hi,
> >
> > AFAIK, none exists.  I am basically the Blackdown JDK for ppc porting 
team
> > and I simply have not had the time to start working on JDK 1.4 yet.  My
> > real need is for help from someone who understands PPC assmebler and
> > instruction set well and can help implement HotSpot for 1.4.
> >
> > IBM has a wonderful JDK with a fast JIT that I have used but it seems 
to be
> > a JDK 1.3.1 version only.  As far as I know they have not even released 
an
> > alpha or beta of JDK 1.4 for ppc linux.
> >
> > Sorry I can't be more help.
> >
> > Kevin
> >
> > On April 21, 2003 07:18 am, florian wrote:
> > > hi!
> > >
> > > does anybody know of a java 1.4 implementation for
> > > ppc? i searched around quite a bit. but didnt find anything..
> > >
> > > anybody knows more?
> > >
> > > ciao!
> > > florian
> > >
> > >
> > > --
> > > To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> > > with a subject of "unsubscribe". Trouble? Contact
> >
> > [EMAIL PROTECTED]
> 
> -- 
> The most subversive thing you can do is assign full humanity to each and 
every 
> individual on this planet.
> 
> 

Re: java 1.4 for ppc

[Kevin B. Hendricks]

Hi,

AFAIK, none exists.  I am basically the Blackdown JDK for ppc porting team 
and I simply have not had the time to start working on JDK 1.4 yet.  My 
real need is for help from someone who understands PPC assmebler and 
instruction set well and can help implement HotSpot for 1.4.

IBM has a wonderful JDK with a fast JIT that I have used but it seems to be 
a JDK 1.3.1 version only.  As far as I know they have not even released an 
alpha or beta of JDK 1.4 for ppc linux.

Sorry I can't be more help.

Kevin

On April 21, 2003 07:18 am, florian wrote:
> hi!
> 
> does anybody know of a java 1.4 implementation for
> ppc? i searched around quite a bit. but didnt find anything..
> 
> anybody knows more?
> 
> ciao!
> florian
> 
> 
> -- 
> To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> with a subject of "unsubscribe". Trouble? Contact 
[EMAIL PROTECTED]
> 
> 

Re: Management of backlight

[Kevin B. Hendricks]

Hi Ben,

> There is currently no solution for external flat panels, I simple don't
> know how backlight is controlled on these.

On a lark I installed pmud and pbbuttonsd but all I ended up with was a 
completely dark screen for my login prompt.  I was blindly able to go to 
/etc/rc.d/init.d/ and do a rm *pmu* and then reboot.

I find it strange that pmud and pbbuttonsd works on an imac and not other 
desktops.  If I remember, the code in the Radeon driver special cases the 
LCD backlight code differently from the DFP.  

Perhaps both are controlled in the same way?  

Is there any similar piece of Darwin source I should be looking at to 
figure this out?

Thanks,

Kevin

Re: Management of backlight

[Kevin B. Hendricks]

Hi,

I have an 17 inch LCD flatpanel and would very much like to control 
brightness and turning off the backlighting after a time period on my 
desktop machine similar to what pbbuttonsd does for PowerBooks.

I tried pbbuttonssd just to see but I get an error message about having no 
/dev/pmu

Anyone with ideas or a similar solution for desktop machines with FlatPanel 
displays?

Kevin



On March 8, 2003 01:46 pm, Matthew Yee-King wrote:
> On Sat, Mar 08, 2003 at 02:10:52PM +0100, David MENTRE wrote:
> > Hello,
> > 
> > On an iBook under MacOS X, I've noticed that MacOS X reduces backlight
> > when the user is inactive and restores backlight as soon as a user 
press
> > a key. Is it possible to have similar behavior under Debian 3.0 ?
> 
> 
> try pbbuttonsd - that can be set up to dim the display just like macosx 
> does.
> 
> cheers
> 
> matthew
> 
> 
> -- 
> To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> with a subject of "unsubscribe". Trouble? Contact 
[EMAIL PROTECTED]
> 
> 

Re: ide-controller ACARD 6260M not working with debian 3.0

[Kevin B. Hendricks]

Hi,
...
> NOTE: the pci-IDE-controller card is named SCSI storage controller above
> !
...
> this is the ide-controller with the hard disk hde connected to, in
> proc/pci it is described as scsi, don't know why.
...

FWIW: I have a similar ACARD and it works fine

01:14.0 SCSI storage controller: Artop Electronic Corp: Unknown device 0009 
(rev 02)
Subsystem: Artop Electronic Corp: Unknown device 0009
Flags: bus master, medium devsel, latency 144, IRQ 54
I/O ports at 0440 [size=8]
I/O ports at 0430 [size=4]
I/O ports at 0420 [size=8]
I/O ports at 0410 [size=4]
I/O ports at 0400 [size=16]
Expansion ROM at 8009 [disabled] [size=64K]
Capabilities: 

AEC6280R: IDE controller on PCI bus 01 dev a0
AEC6280R: chipset revision 2
AEC6280R: not 100% native mode: will probe irqs later
AEC6280R: ROM enabled at 0x8009
ide2: BM-DMA at 0x0400-0x0407, BIOS settings: hde:pio, hdf:pio
ide3: BM-DMA at 0x0408-0x040f, BIOS settings: hdg:pio, hdh:pio

And it also reports itself as a scsi controller under lspci.  I think this 
is done on purpose to allow the card to be seen properly under MacOSX so 
that it can be used for booting from.

It is an IDE card (I have a 60gb ata cdrive attached to it) and as I said 
mine works well.

> > Is this car supported under MacOS? Does it emulate a SCSI controller?

> as you assume the card is supported under MacOS and it emulates a SCSI
> controller as shown in "Apple System Profiler".

I think the answer to this is yes for both the 6260M and the 6280M

> as you assume the card is supported under MacOS and it emulates a SCSI
> controller as shown in "Apple System Profiler".

Are you sure the cable is connected properly (the right type of cable) and 
is the drive capable to be at the speed setting.  Does changing the UDMA 
speed gtet rid of the lost interrupts?

Kevin

Re: Bug#175921: patch: patch works on i386, fails on powerpc

[Kevin B. Hendricks]

Hi,

Sorry I misunderstood the issue.  I thought it could be reproduced under 
ppc Linux.  I did not understand it was a fluke event.

Why would this "fluke" be holding up X11R6 development at all if it could 
not be "reproduced" even by the original submitter?

Sorry, just trying to help.

Kevin

On February 3, 2003 02:05, Michael Fedrowitz wrote:
> On Mon, Feb 03, 2003 at 01:55:34PM -0500, Kevin B. Hendricks wrote:
>
>  Hi,
>
> > Perhaps the code has someplaces where char is explictly assumed to be
> > signed.
>
> Very unlikely, patch is highly portable and I've never heard of any such
> problems with it.
>
> > And then re-run thetest case and see if now passes.
>
> That's just the point: there is no (reproducible) test case. It
> apparently just happened once and silently went away afterwards, which
> is why I don't really believe that it's a bug in patch, but some weird
> intermittent failure of whatever...
>
> -Michael

Re: Bug#175921: patch: patch works on i386, fails on powerpc

[Kevin B. Hendricks]

Hi,

Not sure if this is relevant or not.  The default under PPC Linux gcc is to 
have all char types be unsigned by default whereas on x86 Linux char types 
default to be signed (I believe).  (I belive PPC Linux is unique iin this 
default among Linux platforms)

Perhaps the code has someplaces where char is explictly assumed to be 
signed.

You can easily test this hypothesis by having someone on ppc Linux rebuild 
the patch binary with the following gcc option added:

-fsigned-char

And then re-run thetest case and see if now passes.  If so, simply changing 
the declaration of char to be signed char in the code is enough.

Kevin


On February 3, 2003 01:29, Michael Fedrowitz wrote:
> On Thu, Jan 09, 2003 at 07:35:56PM +0100, Michael Fedrowitz wrote:
>
> [CCing debian-powerpc, see http://bugs.debian.org/175921 for context.
> Anyone here ever seen any ppc specific breakage in GNU patch?]
>
>  Hi,
>
> > Unless this can be reproduced somehow there's probably not a lot
> > anyone can do about it. Otherwise I'd be inclined to blame hardware,
> > kernel, cosmic rays or whatever.
>
> If no one gives me a good reason, why this actually was a bug in patch,
> within a week or two, I'm going to close this.
>
> -Michael

Re: Hardware clock on dual G4 highly inaccurate?

[Kevin B . Hendricks]

Hi,

FWIW on my dual G4 the hardware clock loses between 3 and 5 minutes per 
24 hour period.


I finally set up an hourly cron job to try to keep my clock current.


On Wednesday, January 29, 2003, at 06:35  PM, John P. Fisher wrote:


I have two dual G4s running potato debian.
On both of them the syslog shows time resets from NTP of approximately 
5 seconds per hour. I also noticed independently that the hardware 
clock seemed to drift radically.
Am I misinterpreting? Can the clock be that bad, or should I check 
configuration?

thanks
John




John P. Fisher
at ZNYX Networks
805  683 1488  x 3245
[EMAIL PROTECTED]


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact 
[EMAIL PROTECTED]

recommendation for stats and linear algebra software for PPC Linux

[Kevin B. Hendricks]

Hi,

I need to get back into doing linear algebra and statistics again (basic 
statistics, eigenvectors, regression, principal components, factor 
analysis, structural equation modeling at the level of SAS, SPSS, LISREL, 
etc.

I have found links to netlib and the literally tons of free/public domain 
software available for doing math and statistics.

I spent the afternoon downloading R (the GPL clone of Splus) and Octave, 
and the old EISPack source I remember using once a long time ago, and also 
downloaded blas, atlas, and lapack.

There is alot of stuff out there.   

If I am first and foremost looking for something to replace SAS/stat and 
SPSS primarily for statistics, general linear models, regression instead 
of my now ancient fortran code that invoked EISPACK routines to solve 
linear models,  are these the right packages to be looking at and playing 
with?

If not, what should I be using?  

Ideas and recommendations welcome.

Thanks,

Kevin

Re: gentoo failing to build on powerpc, a bit mysteriously

[Kevin B. Hendricks]

Hi,

I can talking generically about whaqtever project software etc you are 
working on.

Kevin

On August 24, 2002 11:14, Josip Rodin wrote:
> On Sat, Aug 24, 2002 at 10:59:00AM -0400, Kevin B. Hendricks wrote:
> > The key here is to make sure that all shared libraries in the gentoo
> > project are built with -fPIC for all platforms (x86 will work without
> > it but ppc will not but both *should* have -fPIC) .
>
> What gentoo project? I said this is the gentoo PACKAGE. The Debian
> package of the file manager called gentoo.
>
> I thought you people on the debian-powerpc list would realize I'm
> talking about our package if I say it explicitely, and link to our build
> logs about it... :)

Re: gentoo failing to build on powerpc, a bit mysteriously

[Kevin B. Hendricks]

Hi,

There are two possible problem areas:

1. you are missing a required library for linking

or

2.  you are linking to a library that was not properly built with -fPIC

My guess is that the issue is no. 2 since it builds for x86 linux.

The key here is to make sure that all shared libraries in the gentoo 
project are built with -fPIC for all platforms (x86 will work without it 
but ppc will not but both *should* have -fPIC) .

Also look to see if there are any static libraries (lib*.a files) that are 
linked into shared library code.  If so,  those static libraries 
themselves should be built with -fPIC as well since they are destined to 
be used in a shared library.

The linking of lib*.a code built without -fPIC into shared libraries is a 
very common problem / trap that x86 programmers / developers seem to fall 
into since they can't see or experience the failure themselves.

Hoep this gives you some hint as to where to start looking.  If you run 
into trouble, capture the build log to a file and post it someplace and 
send me the link and I will look over it and point out problem areas.

Kevin

On August 24, 2002 10:46, Josip Rodin wrote:
> Hi,
>
> I maintain the gentoo package. No, it doesn't have anything to do with
> the gentoo Linux distribution. :) After having uploaded a new version,
> the powerpc buildd started failing to build gentoo.
>
> http://buildd.debian.org/build.php?pkg=gentoo&arch=powerpc
>
>  mount.o: In function `scan_for_dir':
>  mount.o(.text+0x84): undefined reference to `mne_getmntent'
>  mount.o(.text+0x84): relocation truncated to fit: R_PPC_REL24
> mne_getmntent mount.o: In function `dir_is_mounted':
>  mount.o(.text+0xf0): undefined reference to `mne_setmntent'
>  mount.o(.text+0xf0): relocation truncated to fit: R_PPC_REL24
> mne_setmntent mount.o(.text+0x110): undefined reference to
> `mne_endmntent'
>  mount.o(.text+0x110): relocation truncated to fit: R_PPC_REL24
> mne_endmntent mount.o: In function `mnt_init':
>  mount.o(.text+0x27c): undefined reference to `mne_setmntent'
>  mount.o(.text+0x27c): relocation truncated to fit: R_PPC_REL24
> mne_setmntent mount.o(.text+0x2cc): undefined reference to
> `mne_getmntent'
>  mount.o(.text+0x2cc): relocation truncated to fit: R_PPC_REL24
> mne_getmntent mount.o(.text+0x2dc): undefined reference to
> `mne_endmntent'
>  mount.o(.text+0x2dc): relocation truncated to fit: R_PPC_REL24
> mne_endmntent
>
> I'm not sure how to deal with this, given how it compiles fine
> everywhere else. Can anyone tell me what to do about this?

Re: Java 1.4 for PPC?

[Kevin B. Hendricks]

Hi,

JDK 1.4.X does NOT exist for ppc Linux from Blackdown (or anywhere else 
that I know of yet)/

I am one of the Blackdown ports for ppc Linux so I would know.

If anyone wants a JDK 1.4 from Blackdown (with mozilla plugin) we are going 
to need volunteers who know ppc assembly well.  

So we are waiting for volunteers.  If you know anyone who can help, or 
sponsor someone to do the job, please let me know.

Kevin


On August 12, 2002 09:23, Cesar Cardoso wrote:
> Em Seg, 2002-08-12 às 06:03, Sven LUTHER escreveu:
> > On Sun, Aug 11, 2002 at 08:53:37PM -0700, John Gilger wrote:
> > > To add a twist to the Java plugin thread, is the Java 1.4 SDK or JRE
> > > available anywhere for PPC?
> >
> > I know they are available as .deb for i386, i read it in a mail
> > somewhere, but not if they exist for ppc.
>
> I've download the ppc .tarball, but I heard that there are ppc .debs
> also. Try the Blackdown apt line.

Re: Java 1.4 for PPC?

[Kevin B. Hendricks]

Hi,


No, and it won't be unless IBM does it or Blackdown gets an infusion of 
volunteers that want to help port HotSpot to ppc (people who know ppc 
assembler well).  If you know anyone who wants to help port HotSpot to PPC 
(we begged Apple to release the source to there version but they would 
not) please let me know asap.

Kevin

On August 11, 2002 11:53, John Gilger wrote:
> To add a twist to the Java plugin thread, is the Java 1.4 SDK or JRE
> available anywhere for PPC?
>
> TIA,
>
> John

Re: Mozilla Java Plugin for PPC?

[Kevin B. Hendricks]

Hi,

Checkout Blackdown and hunt throught the mirrors for one that includes the 
PPC binaries and download JDK 1.3.1 and/or jre 1.3.1 and the plugin will 
be part of the installation.

Kevin

On August 9, 2002 10:10, Shawn Dunn wrote:
> Hey all, is there a PPC compatible Java Plugin for Mozilla out there?
> I've checked apt-cache, and done a couple google queries, and no luck
> quite yet, am I just missing something
>
>
>   --Shawn

ANNOUNCING: OpenOffice.org 1.0.1 for PPC Linux

[Kevin B. Hendricks]

Announcing: OpenOffice.org 1.0.1 for PPC Linux

What is it?
--
OpenOffice.org provides a free (source LGPL and no $) almost drop-in 
replacement for Microsoft Office and is available for Solaris (Sparc and 
x86), Linux (x86 and PPC), WIN (NT, 200, XP, 98, 95, etc), and coming real 
soon now (RSN) for FreeBSD, NetBSD, Irix, and Mac OSX.  It provides almost 
transparent file access to Microsoft Office files.  

Please see http://www.openoffice.org for more info and additional download 
sites for major platforms.  PPC Linux specific download sites are provided 
below.


What's New in This Release?
--
- built using gcc-3.1.1 pre 
- improved font handling
- new shared dictionary support
- user-installed hyphenation dictionaries now possible
- huge number of bug fixes and stability improvments

For a more complete listing of fixed bugs:

http://tools.openoffice.org/releases/changelog.html


This version should work on stock Debian, SuSE 7.3, YDL 2.2 and YDL 2.3, 
and almost any other PPC Linux distribution using at least glibc 2.2.1.

Please Note:  Debian users may want to use the official Debian builds which 
include better desktop integration for Debian users and have been built 
and tested on Debian systems (although this version will work too)


DISCLAIMER
--
This is the usual disclaimer.   OOo 1.0.1 should be a rock steady release 
but please don't complain if you loose everything!  I use it constantly 
and have not experienced any problems but I can make no guarantees.
If interested, you can build your own binary direct from source.
(see below).

Please check out http://www.openoffice.org for more info on the 
OpenOffice.org project
 

What's Available (English Only)?
--
- OOo_1.0.1c_LinuxPPC_installer.tar.gz
   md5sum: 432cf242c085fb3a3675779b393ed2b4
  64939379 bytes

- OpenOffice_README.ppclinux

inside the installer is a complete installation/setup
guide in pdf format.


Where can I download it?
---
ftp://ftp.yellowdoglinux.com/pub/yellowdog/software/openoffice/
ftp://ftp.suse/com/pub/projects/powerpc/openoffice/
ftp://ftp.stardiv.de/pub/OpenOffice.org/contrib/linuxppc/

And it is available from up-to-date YellowDog Linux Mirrors:
http://www.yellowdoglinux.com/resources/ftp_mirrors.shtml

Hopefully it will appear on penguinppc.org sites soon.


Building your own from Source
--
Building your own OOo binary is not for the faint hearted and takes at 
least 2 gig of free disk space and up to date tools.  If you are 
interested in building your own version direct from source, please contact 
me and I will try to provide you with and recent or nedded patches and 
some short build instructions.  Right now the build of 1.0.1 requires gcc 
3.X.
 
Thanks,

Kevin B. Hendricks
[EMAIL PROTECTED]



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Re: searching Java & flash for mozilla

[Kevin B . Hendricks]

Hi,

> Is someone hammering the mozilla java maintainer about the
> api issue? It's hard to believe that they are only a few months
> away from declaring mozilla 1.0 and they still can't get the java
> plugin api straight. 

I think they have finally frozen things with mozilla 0.9.7 but we won't 
know until we can build against it and test it with mozilla 0.9.8

> ps If the java plugin hasn't been built with -fPIC as it should
> have been in the past, I wonder if this change will eliminate
> any other odd behavior we had been seeing in the past.

No when I built it, it was always built with -fPIC.   I simply had not 
commited all of my changes to the tree (I missed some changes to makefiles 
by accident) so his build did not properly pick it up.

Kevin

Re: searching Java & flash for mozilla

[Kevin B . Hendricks]

Hi Michel,

> It is indeed strange that the JDK plugin doesn't even have a relocation
> for 'ect' - could it be a bug in the dynamic linker after all, mangling
> the relocation for 'select' or another symbol containing 'ect'?

Yes, I think you are absolutely right.  It was probably some other symbol 
containing the substring "ect".  We should probably file a bug report on 
the linker error message for this.

But forcing -fPIC throughout the jdk plugin build seems to have done the 
trick.  Laurent can now run Galeon properlly with his latest build.

Unfortunately, you can not even build the jdk plugin against mozilla 0.9.7 
since the jdk plugin code and the mozilla headers are out of sync.  The 
x86 JDK crowd built against mozilla 0.9.5 instead of 0.9.7 and it seems to 
work.

So look for a new ppc linux release from Laurent fixing that problem later 
in the week (the mirrors have to get updated). 

Thanks for your bug reports (and hints!).

Kevin

Re: searching Java & flash for mozilla

[Kevin B . Hendricks]

Hi, 
I am no Galeon user but Laurent  just reported the following:

> just double-checked. The plugin is built with fPIC option and it's a
> shared lib all right. There is no ect symbol per se in the plugin code.
> Mozilla loads applets correctly with the plugin and they run well. And
> Galeon 1.0.2 shows the same error on my machine. Weird.

So the "ect" missing symbol is not part of the jdk plugin?  Could one of 
the Galeon libraries be missing a -fPIC in its build? 

Ideas here?

Kevin

On January 26, 2002 11:50, Michel Dänzer wrote:
> On Sam, 2002-01-26 at 02:16, Brian Victor wrote:
> > On Fri, Jan 25, 2002 at 05:02:59PM -0800, Wilhelm *Rafial* Fitzpatrick 
wrote:
> > >At 4:45 PM -0500 1/25/02, Adam Goode wrote:
> > >>Java can be found (in various forms) as part of Debian, and as part
> > >> of the Blackdown project
> > >> (http://www.blackdown.org/java-linux.html). The Blackdown version
> > >> contains Sun's Java plugin which should work on Mozilla.
> > >
> > >Sadly, the PPC java plugin (which can be found at
> > >http://penguinppc.org/projects/java) broke a couple Mozilla revs
> > >(around 9.5 or so) back, and no one to my knowledge has managed to
> > >release an updated version.
> > >
> > >If I am incorrect about this, please let me know, I sorely miss Java
> > >when surfing with Mozilla.
> >
> > Actually, blackdown released 1.3.1-02a-FCS three days ago.  I
> > installed it as soon as I found out and java has been working about as
> > well as I can expect in Mozilla 0.9.7.  (i.e., most applets work
> > pretty well, a few like to munch processor time eternally.)  See the
> > address Adam gave for details.
>
> Thanks, but sadly, it gives this error with galeon:
>
> LoadPlugin: failed to initialize shared library
> /usr/local/j2sdk1.3.1/jre/plugin/ppc/mozilla/javaplugin_oji.so
> [/usr/local/j2sdk1.3.1/jre/plugin/ppc/mozilla/javaplugin_oji.so:
> R_PPC_REL24 relocation at 0x0d63d1c4 for symbol `ect' out of range]
>
> It seems the plugin contains non-PIC code. Kevin, do you still build the
> PPC binaries, or where should I report this?

Re: searching Java & flash for mozilla

[Kevin B . Hendricks]

Hi,

No, I did not build this binary and probably won't make future builds 
either due to time constraints.  But I do not understand how the plugin 
can link fine under mozilla and not galeon?

I will pass along your bug report to Blackdown internal porting java-linux 
mailing list.  There is a Bug reporting system at 
http://java.blackdown.org as well.

Thanks,

Kevin


On January 26, 2002 11:50, Michel Dänzer wrote:
> On Sam, 2002-01-26 at 02:16, Brian Victor wrote:
> > On Fri, Jan 25, 2002 at 05:02:59PM -0800, Wilhelm *Rafial* Fitzpatrick 
wrote:
> > >At 4:45 PM -0500 1/25/02, Adam Goode wrote:
> > >>Java can be found (in various forms) as part of Debian, and as part
> > >> of the Blackdown project
> > >> (http://www.blackdown.org/java-linux.html). The Blackdown version
> > >> contains Sun's Java plugin which should work on Mozilla.
> > >
> > >Sadly, the PPC java plugin (which can be found at
> > >http://penguinppc.org/projects/java) broke a couple Mozilla revs
> > >(around 9.5 or so) back, and no one to my knowledge has managed to
> > >release an updated version.
> > >
> > >If I am incorrect about this, please let me know, I sorely miss Java
> > >when surfing with Mozilla.
> >
> > Actually, blackdown released 1.3.1-02a-FCS three days ago.  I
> > installed it as soon as I found out and java has been working about as
> > well as I can expect in Mozilla 0.9.7.  (i.e., most applets work
> > pretty well, a few like to munch processor time eternally.)  See the
> > address Adam gave for details.
>
> Thanks, but sadly, it gives this error with galeon:
>
> LoadPlugin: failed to initialize shared library
> /usr/local/j2sdk1.3.1/jre/plugin/ppc/mozilla/javaplugin_oji.so
> [/usr/local/j2sdk1.3.1/jre/plugin/ppc/mozilla/javaplugin_oji.so:
> R_PPC_REL24 relocation at 0x0d63d1c4 for symbol `ect' out of range]
>
> It seems the plugin contains non-PIC code. Kevin, do you still build the
> PPC binaries, or where should I report this?

Re: Bug#86356: analog: analog segfaults

[Kevin B . Hendricks]

Hi,

Regardless of the interface or code, the compiler should be able to handle 
this case properly.  I agree simpler is better but I am more interested in 
making sure this bug is fixed in gcc if it hasn't already been.

 If not, I personally want to see it fixed since the code I wrote for 
sys_invokeNative in the JDK and the openoffice bridges code and the Mozilla 
code, and the libffi code, and ... are wrong and need to be changed since 
they all follow the published abi.

Kevin

On Friday 23 February 2001 17:10, Andrew Sharp wrote:
> One look at that interface to printtree is all that is needed to see
> where the real problem is.  Whoever wrote this code is badly in need
> of a long and meaningful "timeout" with _The Elements of Programming
> Style_ by Kernighan & Plauger.  KISS.  Geez, build a structure and
> pass the pointer, rather than the much slower [and apparently
> bugier, and painful to read] method of trying to force the compiler
> and arg passing code to deal with that mountain of 
> 
> a
>   
> "Kevin B. Hendricks" wrote:
> > 
> > Hi,
> > 
> > I think the second double value is confusing the compiler into skipping a
> > stack slot when it really shouldn't be doing that at all!
> > 
> > This is wierd.
> > 
> > Here is a quick and dirty way to test.  Move both double parameters to the
> > beginning of the function and caller and the problem should go away.
> > 
> > Another solution is to include a "dummy" int variable in both the caller
> > and the function right before the double parameter "unit".  That dummy 
will
> > fill a stack slot and force any messed up double alignment issue to become
> > moot.
> > 
> > If either of those workarounds work, then please pass all of this info to
> > Franz Sirl's attention on the gcc@gcc.gnu.org site and he can use it to
> > track down the messed up code. It the workarounds fix things, this is a
> > definite bug
> > 
> > Okay, here is what should be where:
> > 
> > gpr registers
> > r3   outf
> > r4   rep
> > r5   outstyle
> > r6   multibyte
> > r7   tree
> > r8   requests
> > r9   date
> > r10  badp
> > 
> > floating point registers
> > f1  totb
> > f2  unit
> > f3
> > f4
> > f5
> > f6
> > f7
> > f8
> > 
> > overflow stack (starts aligned to 8 at the previous frame pointer + 8
> > offset  0: badn
> > offset  4: level
> > offset  8: partname
> > offset  c: aliashead
> > offset 10: linkhead
> > offset 14: baseurl
> > offset 18: totr
> > offset 1c: totp
> > offset 20: width
> > offset 24: possrightalign
> > offset 28: bmult
> >  <== (if passed on the stack the double would have
> >   been here but there were enough floating point
> >   registers so it should not be on the stack.)
> >   (However, if it was on the stack, the compiler 
should
> >have skipped a stack slot since doubles must be
> >passed aligned to 8)
> > offset 2c: sepchar
> > offset 30: rsepchar
> > offset 34: decpt
> > offset 38: compsep
> > offset 3c: rawbytes
> > offset 40: cols
> > offset 44: colhead
> > offset 48: colheadp
> > offset 4c: gender
> > offset 50: html
> > offset 54: monthname
> > offset 58: dayname
> > offset 5c: monthlen
> > offset 60: daylen
> > offset 64: plainmonthend
> > offset 68: plaindaylen
> > offset 6c: lngstr
> > 
> > Please let me know if the workaround  "fixes" things.  We will then have a
> > bug.
> > 
> > Thanks,
> > 
> > Kevin
> > 
> > On Friday 23 February 2001 15:46, Stephen Turner wrote:
> > > Thanks for your help with this, Kevin (I'm the upstream author).
> > >
> > > > To see if it is indeed a parameter passing issue, I need to know what 
the
> > > > types are for each parameter passed below (specifically if any are 
long
> > > > long int or float or double types and what the return type is of that
> > > > function so that I can tell is any structures are returned.
> > > >
> > >
> > > The definition:
> > >
> > > typedef unsigned char logical;
> > > typedef signed char choice;
> > > /* and Strlist, Alias, Include are typedefs to structs */
> > > void printtree(FILE *outf, choice rep, choice outstyle, logical 
multibyte,
> > >Hashtable *tree,

Re: Bug#86356: analog: analog segfaults

[Kevin B . Hendricks]

Hi,

I have a working gcc HEAD build from about 2 weeks ago.  If you send me some 
standalone test code, I would be happy to test it.

I also have 2.95.3 too and will test with both.

Just create a main with a call to printtrace and have printtrace print all 
the values and just return and I will test it for you.

Take care,

Kevin

On Friday 23 February 2001 16:59, Stephen Turner wrote:
> On Fri, 23 Feb 2001, Kevin B. Hendricks wrote:
> > 
> > Here is a quick and dirty way to test.  Move both double parameters to 
the 
> > beginning of the function and caller and the problem should go away.
> > 
> > Another solution is to include a "dummy" int variable in both the caller
> > and the function right before the double parameter "unit".  That dummy 
will
> > fill a stack slot and force any messed up double alignment issue to become
> > moot.
> > 
> 
> The second fix got it past the call to printtree(). Then it crashed when
> calling another function, printcols(), which I fixed with the first fix.
> This allowed it to run without crashing, but the resultant output was
> obviously wrong, with what could have been a related bug.
> 
> Anyway, I think this proves that your hypothesis was correct.
> 
> > If either of those workarounds work, then please pass all of this info to
> > Franz Sirl's attention on the gcc@gcc.gnu.org site and he can use it to
> > track down the messed up code. It the workarounds fix things, this is a 
> > definite bug
> > 
> 
> You said these were mostly fixed in the 2.95.3 series. The original bug
> filer is using 2.95.2. Should I still file a bug? Or has someone got a
> nightly build or something that they could test it on first, in case it's
> already been fixed?
> 
> -- 
> Stephen Turner   http://www.statslab.cam.ac.uk/~sret1/
>   Statistical Laboratory, Wilberforce Road, Cambridge, CB3 0WB, England
>   "Your account can only be used for a single internet session at any one
>time and for no more than 24 hours in any one day." (NTL terms of use)

Re: Bug#86356: analog: analog segfaults

[Kevin B . Hendricks]

Hi,

I think the second double value is confusing the compiler into skipping a 
stack slot when it really shouldn't be doing that at all!

This is wierd.

Here is a quick and dirty way to test.  Move both double parameters to the 
beginning of the function and caller and the problem should go away.

Another solution is to include a "dummy" int variable in both the caller
and the function right before the double parameter "unit".  That dummy will
fill a stack slot and force any messed up double alignment issue to become
moot.

If either of those workarounds work, then please pass all of this info to
Franz Sirl's attention on the gcc@gcc.gnu.org site and he can use it to
track down the messed up code. It the workarounds fix things, this is a 
definite bug

Okay, here is what should be where:

gpr registers
r3   outf
r4   rep
r5   outstyle
r6   multibyte
r7   tree
r8   requests
r9   date
r10  badp


floating point registers
f1  totb
f2  unit
f3
f4
f5
f6
f7
f8


overflow stack (starts aligned to 8 at the previous frame pointer + 8
offset  0: badn
offset  4: level
offset  8: partname
offset  c: aliashead
offset 10: linkhead
offset 14: baseurl
offset 18: totr
offset 1c: totp
offset 20: width
offset 24: possrightalign
offset 28: bmult
 <== (if passed on the stack the double would have 
  been here but there were enough floating point
  registers so it should not be on the stack.)
  (However, if it was on the stack, the compiler should
   have skipped a stack slot since doubles must be
   passed aligned to 8)
offset 2c: sepchar
offset 30: rsepchar
offset 34: decpt
offset 38: compsep
offset 3c: rawbytes
offset 40: cols
offset 44: colhead
offset 48: colheadp
offset 4c: gender
offset 50: html
offset 54: monthname
offset 58: dayname
offset 5c: monthlen
offset 60: daylen
offset 64: plainmonthend
offset 68: plaindaylen
offset 6c: lngstr



Please let me know if the workaround  "fixes" things.  We will then have a 
bug.

Thanks,

Kevin

On Friday 23 February 2001 15:46, Stephen Turner wrote:
> Thanks for your help with this, Kevin (I'm the upstream author).
> 
> > To see if it is indeed a parameter passing issue, I need to know what the 
> > types are for each parameter passed below (specifically if any are long
> > long int or float or double types and what the return type is of that
> > function so that I can tell is any structures are returned.
> >
> 
> The definition:
> 
> typedef unsigned char logical;
> typedef signed char choice;
> /* and Strlist, Alias, Include are typedefs to structs */
> void printtree(FILE *outf, choice rep, choice outstyle, logical multibyte,
>Hashtable *tree, choice requests, choice date, Hashentry *badp,
>unsigned long badn, unsigned int level, Strlist *partname,
>Alias *aliashead, Include *linkhead, char *baseurl,
>unsigned long totr, unsigned long totp, double totb,
>unsigned int width[], logical possrightalign,
>unsigned int bmult, double unit, char sepchar, char repsepchar,
>char decpt, char *compsep, logical rawbytes, choice *cols,
>char *colhead, char *colheadp, char gender, logical *html,
>char **monthname, char **dayname, unsigned int monthlen,
>unsigned int daylen, unsigned int plainmonthlen,
>unsigned int plaindaylen, char **lngstr) {
> 
> The call:
> 
> printtree(outf, rep, outstyle, multibyte, tree, requests, date, badp, badn,
> 0, NULL, aliashead, linkhead, baseurl, totr, totp, totb, width,
> possrightalign, bmult, unit, sepchar, repsepchar, decpt, compsep,
> rawbytes, cols, colhead, colheadp, gender, html, monthname,
> dayname, monthlen, daylen, plainmonthlen, plaindaylen, lngstr);
> 
> I've double-checked that all arguments in the call have the correct types.
> 
> However, notice that printtree() has 38 arguments. The C standard (Section
> 5.2.4.1) only requires implementations to accept 31 arguments. Does gcc have
> this limit?
> 
> > Another (easier solution) is to modify each routine to print the values 
of 
> > all parameters just before the call and just inside the called routine.
> 
> I've done this. fprintf'ing the values of all the parameters immediately
> before the call and immediately on entry to the function gives:
> 
> BEFORE:
> 0x100f3f48 9 0 0 0x1007f550 0 4 0xffe859c
> 268919984 0 (nil) (nil) 0x100e8498 (nil)
> 1 0 88140.00 0x78f8 0 0
> 1.00 44 0 46 0x1007e498 0 0x100654de
> 0x100e9eb8 0x100e9ec8 n 0x1006543f 0x1006592c 0x10065910
> 3 3 3 3 0x100e98b0
> 
> AFTER:
> 0x100f3f48 9 0 0 0x1007f550 0 4 0xffe859c
> 268919984 0 (nil) (nil) 0x100e8498 (nil)
> 1 0 88140.00 0x78f8 0 0
> 1.00 0 46 152 (nil) 222 0x100e9eb8
> 0x100e9ec8 0x6e ? 0x1006592c 0x10065910 0x3
> 3 3 3 269392048 0x100f3f48
> Segmentation fault
> 
> Notice how the second half of the arguments appear to have been shifted up
> one. Compare with the same code 

Re: [S.R.E.Turner@statslab.cam.ac.uk: Re: Bug#86356: analog: analog segfaults]

[Kevin B . Hendricks]

Hi,

Most of the bugs in passing parameters have been fixed by the 2.95.3 series 
Franz Sirl has created.

To see if it is indeed a parameter passing issue, I need to know what the 
types are for each parameter passed below (specifically if any are long long 
int or float or double types and what the return type is of that function so 
that I can tell is any structures are returned.

With that info, I can tell you which registers should have which values on 
entry to printtree, and what the contents of the overflow stack should look 
like.

You can then set a breakpoint just inside the printtree frame and do an info 
registers and an x/30x on the previous frames stack pointer (where the 
outgoing overflow parameters are actually stored) to check the remaining 
parameters.

Another (easier solution) is to modify each routine to print the values of 
all parameters just before the call and just inside the called routine.

That will tell you if gcc miscompiled anything.

Kevin
 

On Thursday 22 February 2001 01:55, Michel Lanners wrote:
> [cross-posted to linuxppc-dev]
> 
> On  20 Feb, this message from [EMAIL PROTECTED] echoed through cyberspace:
> > Background: analog segfaults when using a custom config file.
> > The author tried debugging the program on my B50 and replied with this
> > message.
> > Is gcc known to be buggy for powerpc? I'm running 2.95.2-17 and the
> > program has been compiled without optimizations.
> 
> No idea whether it _is_ buggy or not. However, the list of arguments to
> that function below seems rather long to me, so if it is a bug in gcc,
> it might not have been noticed before.
> 
> Anyway, one thing to look at might be the fact that not all function
> arguments are passed the same way according to the ABI we are using. The
> first few are passed in registers, the rest on the stack (IIRC).
> However, I don't know exactly how many are passed in registers. But it
> would seem that 22 (the arguments coming through OK below) is a bit high
> a number...
> 
> Any of the gcc gurus around?
> 
> Michel
> 
> > - Forwarded message from Stephen Turner 
<[EMAIL PROTECTED]> -
> >
> > OK, I've discovered what's going on, but I have no idea WHY it's going on.
> >
> > The problem comes at line 736 of output.c. That line calls printtree, as
> > follows:
> >
> > printtree(outf, rep, outstyle, multibyte, tree, requests, date, badp, 
badn,
> >   0, NULL, aliashead, linkhead, baseurl, totr, totp, totb, width,
> >   possrightalign, bmult, unit, sepchar, repsepchar, decpt, 
compsep,
> >   rawbytes, cols, colhead, colheadp, gender, html, monthname,
> >   dayname, monthlen, daylen, plainmonthlen, plaindaylen, lngstr);
> >
> > But by the time we reach printtree, several of the values are wrong.
> > Everything is fine up to unit. But sepchar got repsepchar's true value 
(0),
> > repsepchar got decpt's true value ('.'), etc., until plaindaylen got
> > lngstr's value (0x100e93c0 = 269390784) and lngstr got a random value.
> >
> > As I say, I don't really see how this can have happened. Have you
> > encountered anything like this before?
> >
> > - End forwarded message -
> 
> 
> -
> Michel Lanners |  " Read Philosophy.  Study Art.
> 23, Rue Paul Henkes|Ask Questions.  Make Mistakes.
> L-1710 Luxembourg  |
> email   [EMAIL PROTECTED]|
> http://www.cpu.lu/~mlan| Learn Always. "
> 
> ** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: [ANN] JDK 1.3.0 Beta B2, JDK 1.2.2 FCS and JDK 1.1.8 V3 for PPP Linux

[Kevin B. Hendricks]

Whoops!

Nothing like posting the wrong url!

Sorry about that.  I can't seem to type at all today (notice the subject
line with its PPP Linux and not PPC Linux).


As Jeramy correctly posted, the correct URL is

http://linuxppc.org/usr/java

Oh well!

Kevin


> > http://linuxppc.org/user/java
> >
> >
> 
> http://linuxppc.org/usr/java
> 
> Have Fun!
> 
> -Jeramy

[ANN] JDK 1.3.0 Beta B2, JDK 1.2.2 FCS and JDK 1.1.8 V3 for PPP Linux

[Kevin B. Hendricks]

ANNOUNCING:

  JDK 1.1.8 FCS, JDK 1.2.2 FCS, and JDK 1.3.0 Beta 2
  For PPC Linux distributions


Blackdown Organization is pleased to announce the availability
of JDK 1.1.8, JDK 1.2.2 and JDK 1.3.0 Beta 2 for current PPC Linux
distributions.


See the new for Java on PPC Linux Website at:  

http://linuxppc.org/user/java 


More mirror sites (including the official Blackdown sites) will be 
coming along later (hopefully!).

If anyone else wants to mirror this site, please do and pass
along your URL to me, please.


Notes and Requirements
--
- should work with any *current* distribution:
 YDL 1.2, LinuxPPC 2000, SuSE 6.4, Debian, etc

- lots of bug fixes from both Sun and Blackdown

- still no JIT for JDK 1.2.2 or JDK 1.3.0 (we need your help!)

- requires glibc 2.1.3 or higher (do *not* use glibc-2.1.3-15c.ppc.rpm set)

- requires XFree 3.6 or higher or Xpmac rev10 or higher

- requires a recent version of tar that supports bzip2 

- requires dynamic OpenMotif 2.1.30-3 (with an additional JDK bugfix patch)


Check out the website for more INFO in the Release Notes

Hope this Helps.

Kevin B. Hendricks

ps. Thank Jeremy for getting the new linuxppc.org site off the ground and
hosting 
the new java on ppc linux home!