Re: 64bit version and running 32bit on a 64bit system

2009-10-27 Thread Alexander Burger
On Mon, Oct 26, 2009 at 04:46:33PM +0100, Henrik Sarvell wrote:
 About the Mac stuff, AFAIK everything works OK on the Intel architecture.
 
 Putting even a joule of effort into making things work on the old
 Motorola processors would be too much.

The problem on the Mac is not the CPU, but the development system
or runtime library.
-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe


Re: 64bit version and running 32bit on a 64bit system

2009-10-27 Thread Alexander Burger
On Tue, Oct 27, 2009 at 02:56:49AM +0100, Henrik Sarvell wrote:
 (in opml.xml (and (xml?) (xml)))
 
 That's how far (traceAll) goes, it simply manages to output 'xml'

Great! That was it. I traced a little more into 'xml', and found that
'till' was the culprit. Forgot to save/restore the 'Z' register, stupid
mistake. Uploaded a new test release.

Please find more of these!

Thanks,
- Alex
-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe


Re: 64bit version and running 32bit on a 64bit system

2009-10-26 Thread Alexander Burger
Hi Henrik,

 idea what might be the problem, it just crashes without any errors and
 churns away with one process on each cpu, eating 100% on both.

Hm, what might be the reason? Can you try trace it to find you where
exactly it hangs?


 However when I try to run the 32bit server on my 64bit system i now
 get ht:Pack -- Undefined, how come?

You probably have to make (cd src; make clean) to remove the lib/ht
library before you 'make'.

Cheers,
- Alex
-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe


Re: 64bit version and running 32bit on a 64bit system

2009-10-26 Thread Henrik Sarvell
What flags do I need? If I recall correctly I think I tried compiling
the 32bit version but it didn't work simply using 'make'.



On Mon, Oct 26, 2009 at 7:05 AM, Alexander Burger a...@software-lab.de wrote:
 Hi Henrik,

 idea what might be the problem, it just crashes without any errors and
 churns away with one process on each cpu, eating 100% on both.

 Hm, what might be the reason? Can you try trace it to find you where
 exactly it hangs?


 However when I try to run the 32bit server on my 64bit system i now
 get ht:Pack -- Undefined, how come?

 You probably have to make (cd src; make clean) to remove the lib/ht
 library before you 'make'.

 Cheers,
 - Alex
 --
 UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe

-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe


Re: 64bit version and running 32bit on a 64bit system

2009-10-26 Thread Alexander Burger
On Mon, Oct 26, 2009 at 12:09:59PM +0100, Henrik Sarvell wrote:
 What flags do I need? If I recall correctly I think I tried compiling
 the 32bit version but it didn't work simply using 'make'.

Compiler flags? They should be all in Makefile, so there is just a

   (cd src; make)


  However when I try to run the 32bit server on my 64bit system i now
  get ht:Pack -- Undefined, how come?
 
  You probably have to make (cd src; make clean) to remove the lib/ht
  library before you 'make'.

I believe that because you've first built the 64-bit version, there is
already a library lib/ht, so this could be the reason why you got

   ht:Pack -- Undefined



BTW, it would be very helpful for me if you could find out the reason
why your import didn't work. Otherwise, I cannot fix that bug. For a
first try, I would say

   (traceAll)

before you start your import, and see how far it goes. You can probably
see then which function is called but doesn't return. In the second
step, you could then explicitly also trace some built-in functions
called by that function (as 'traceAll' traces only Lisp-level
functions).


I'd really appreciate if people could send me precise bug reports, after
investigating the reason of a problem. Not just it doesn't work or it
crashes. I'd like to know exactly which functions goes wrong. We have
the same situation with the Mac users. We know that something goes wrong
when loading shared object files, but nobody cares to take a closer look
with a debugger. Quit disappointing ...

Cheers,
- Alex
-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe


Re: 64bit version and running 32bit on a 64bit system

2009-10-26 Thread Alexander Burger
Hi Henrik,

 When it comes to ht:Pack -- Undefined, no I hadn't built the 64bit
 first, I simply used the old version I had already compiled on a 32bit
 system, maybe that's why it didn't work?

An Undefined error is issued, when

   - the name of an undefined symbol does not contain a colon,

   - the library named by the part before the colon cannot be found,

   - or a function named by the part after the colon cannot be located
 in that library.

If the 'ht' library exists, I would expect it to contain 'Pack'. Is it
possible that the shared object file lib/ht somehow got lost, or was
not built with (cd src; make)?

Cheers,
- Alex
-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe


Re: 64bit version and running 32bit on a 64bit system

2009-10-26 Thread Henrik Sarvell
About the Mac stuff, AFAIK everything works OK on the Intel architecture.

Putting even a joule of effort into making things work on the old
Motorola processors would be too much.


On Mon, Oct 26, 2009 at 2:46 PM, Alexander Burger a...@software-lab.de wro=
te:
 On Mon, Oct 26, 2009 at 02:11:41PM +0100, Henrik Sarvell wrote:
 The only thing that is different is that I'm now running them on a
 64bit system. The interpreter works just fine, it's just that ht:Pack
 thing that won't play ball.

 Strange. What does 'file' say?

 =A0 $ file lib/ht
 =A0 lib/ht: ELF 32-bit LSB shared object, Intel 80386, version 1 (SYSV), =
dynamically linked, stripped

 Looking in lib/ht should show that 'Pack' exists:

 =A0 $ strings lib/ht |fgrep Pack
 =A0 Pack

 Does this the case? Perhaps we can look at it together later in IRC?


 Oh btw, when it comes to the 64bit crash, I noticed that the database
 files were not numbered anymore but had file names such as for
 instance '@', and '@A', I suppose that's as it should be?

 Yes, that's correct. The internal structure, the names of the DB files,
 and the names of external symbols changed.

 Files now have names consisting of '@' and the letters 'A' through 'O'.

 External symbols no longer have a hyphen in their name, and consist of a
 sequence of the above letters for the file part, and octal digits for
 the object part (e.g. {AC71231}).

 Cheers,
 - Alex
 --
 UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=3dunsubscribe

-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe


Re: 64bit version and running 32bit on a 64bit system

2009-10-26 Thread Henrik Sarvell
Seems like I have to apologize for the 32bit confusion, when I tried
today it simply worked, no ht:Pack problems.

Anyway, I ran my import script which starts off with trying to load an xml =
file:

(in opml.xml (and (xml?) (xml)))

That's how far (traceAll) goes, it simply manages to output 'xml'
before the interpreter locks up completely at 100% of cpu, I have to
kill it every time.



On Mon, Oct 26, 2009 at 4:46 PM, Henrik Sarvell hsarv...@gmail.com wrote:
 About the Mac stuff, AFAIK everything works OK on the Intel architecture.

 Putting even a joule of effort into making things work on the old
 Motorola processors would be too much.


 On Mon, Oct 26, 2009 at 2:46 PM, Alexander Burger a...@software-lab.de w=
rote:
 On Mon, Oct 26, 2009 at 02:11:41PM +0100, Henrik Sarvell wrote:
 The only thing that is different is that I'm now running them on a
 64bit system. The interpreter works just fine, it's just that ht:Pack
 thing that won't play ball.

 Strange. What does 'file' say?

 =A0 $ file lib/ht
 =A0 lib/ht: ELF 32-bit LSB shared object, Intel 80386, version 1 (SYSV),=
 dynamically linked, stripped

 Looking in lib/ht should show that 'Pack' exists:

 =A0 $ strings lib/ht |fgrep Pack
 =A0 Pack

 Does this the case? Perhaps we can look at it together later in IRC?


 Oh btw, when it comes to the 64bit crash, I noticed that the database
 files were not numbered anymore but had file names such as for
 instance '@', and '@A', I suppose that's as it should be?

 Yes, that's correct. The internal structure, the names of the DB files,
 and the names of external symbols changed.

 Files now have names consisting of '@' and the letters 'A' through 'O'.

 External symbols no longer have a hyphen in their name, and consist of a
 sequence of the above letters for the file part, and octal digits for
 the object part (e.g. {AC71231}).

 Cheers,
 - Alex
 --
 UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=3dunsubscribe


-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe


Re: 64bit version and running 32bit on a 64bit system

2009-10-26 Thread Henrik Sarvell
The last part of my earlier post refers to the 64bit version, forgot
to say that.


On Tue, Oct 27, 2009 at 2:56 AM, Henrik Sarvell hsarv...@gmail.com wrote:
 Seems like I have to apologize for the 32bit confusion, when I tried
 today it simply worked, no ht:Pack problems.

 Anyway, I ran my import script which starts off with trying to load an xm=
l file:

 (in opml.xml (and (xml?) (xml)))

 That's how far (traceAll) goes, it simply manages to output 'xml'
 before the interpreter locks up completely at 100% of cpu, I have to
 kill it every time.



 On Mon, Oct 26, 2009 at 4:46 PM, Henrik Sarvell hsarv...@gmail.com wrot=
e:
 About the Mac stuff, AFAIK everything works OK on the Intel architecture=