Re: [expert] kernel bug still not fixed
Viestissä Maanantai 7. Heinäkuuta 2003 05:53, Glenn Burkhardt kirjoitti: I've built a stock 2.4.21 kernel, and applied the patch SGI XFS snapshot-2.4.21-2003-06-23_01:45_UTC and I can run the debugger. There are still a bunch of drivers to get straight, if I want to use it (which is why I prefer getting a kernel from Mandrake). This is 1.3 release. There is no patch for the 1.2 release for the 2.4.21 kernel; the only one is for the 2.4.19 kernel. No it's not... It's a 1.3 pre2 snapshot Which makes me wonder where the code in the current 9.1 releases came from. They all (including cooker linux-2.4.21-2mdk) have the version: snapshot 2.4.20-2003-01-14_00:43_UTC The official 1.2 release was made 2003/2/11. So was this an early release of 1.2 or 1.3, and how was the patch made compatible with the 2.4.21 kernel, since it was picked up a month before XFS 1.2 was released, Back in January the above snapshot was stated to become 1.2 Release I also verifyed it then when 1.2 got Released, and the code was the same.. which is, at least now, only available for the 2.4.19 kernel. You can still find the snapshots for 2.4.20 and a lot of other kernels at: ftp://oss.sgi.com/projects/xfs/download/patches If IRC the 2.4.19 XFS 1.2 patches applies cleanly onto 2.4.20, so there is no need to keep duplicates... or maybe SGI knows of this promblem with 2.4.20, and therefore skipped the 1.2 Release for 2.4.20... And with 1.3 around the corner... who cares about the old stuff... ;-) It would at least be sporting if Mandrake would post a note on the Errata page that 9.1 shouldn't use XFS if the system is to be used for development. Apparently it's only the root partition that matters. I'm in the process of reinstalling both 9.1 and Cooker on my system, this time with everything on XFS, so I'll will track down thwe bug if/when it appears for me... Also, there's no mention of problems with 'gdb' on any of the XFS mailing lists or change logs, so my guess is that a patch for an older version of the kernel was force fitted to the 2.4.21 kernel, and this was an edge that got shaved off. well actually MDK 9.1 does not have a full 2.4.21 kernel (since it was released 2003-06-13), only a 2.4.21pre based + patches from some of the rc's + a lot of MDK addons... One thing is for sure (unless SGI drags their feets...) for the upcoming MDK 9.2 there will be a new shiny XFS 1.3 added AFAIK... Regards Thomas Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
I've built a stock 2.4.21 kernel, and applied the patch SGI XFS snapshot-2.4.21-2003-06-23_01:45_UTC and I can run the debugger. There are still a bunch of drivers to get straight, if I want to use it (which is why I prefer getting a kernel from Mandrake). This is 1.3 release. There is no patch for the 1.2 release for the 2.4.21 kernel; the only one is for the 2.4.19 kernel. Which makes me wonder where the code in the current 9.1 releases came from. They all (including cooker linux-2.4.21-2mdk) have the version: snapshot 2.4.20-2003-01-14_00:43_UTC The official 1.2 release was made 2003/2/11. So was this an early release of 1.2 or 1.3, and how was the patch made compatible with the 2.4.21 kernel, since it was picked up a month before XFS 1.2 was released, which is, at least now, only available for the 2.4.19 kernel. It would at least be sporting if Mandrake would post a note on the Errata page that 9.1 shouldn't use XFS if the system is to be used for development. Apparently it's only the root partition that matters. Also, there's no mention of problems with 'gdb' on any of the XFS mailing lists or change logs, so my guess is that a patch for an older version of the kernel was force fitted to the 2.4.21 kernel, and this was an edge that got shaved off. Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
On Friday 04 July 2003 03:46 am, you wrote: No. The problem has not been fixed. I keep emailing mandrake about the problem, but they don't seem to care. I think they need to pull some new XFS patches from the SGI guys and see if that works. The SGI kernel doesn't have this problem. Nor does any of the 9.0 kernels that mandrake has produced. Also, the cooker kernels seem to have the same problem. Thanks a lot for the additional info! It probably wouldn't have dawned on my that the only system I've built with mdk 9.1 on which the debugger works is the one that uses ext3. The other 3 use xfs! Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
No. The problem has not been fixed. I keep emailing mandrake about the problem, but they don't seem to care. I think they need to pull some new XFS patches from the SGI guys and see if that works. The SGI kernel doesn't have this problem. Nor does any of the 9.0 kernels that mandrake has produced. Also, the cooker kernels seem to have the same problem. Thanks a lot for the additional info! It probably wouldn't have dawned on my that the only system I've built with mdk 9.1 on which the debugger works is the one that uses ext3. The other 3 use xfs! Why not simply get a GNU or Redhat kernel? Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
From: Jim C [EMAIL PROTECTED] No. The problem has not been fixed. I keep emailing mandrake about the problem, but they don't seem to care. I think they need to pull some new XFS patches from the SGI guys and see if that works. The SGI kernel doesn't have this problem. Nor does any of the 9.0 kernels that mandrake has produced. Also, the cooker kernels seem to have the same problem. I don't think the XFS will be updated until XFS 1.3 is officially released... (should happend soon AFAIK...) Thanks a lot for the additional info! It probably wouldn't have dawned on my that the only system I've built with mdk 9.1 on which the debugger works is the one that uses ext3. The other 3 use xfs! Why not simply get a GNU or Redhat kernel? Better to get it fixed in MDK kernel ;-) I'll put it on my TODO list... Regards Thomas Backlund [EMAIL PROTECTED] Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
On Friday 04 July 2003 12:16 pm, Jim C wrote: No. The problem has not been fixed. I keep emailing mandrake about the problem, but they don't seem to care. I think they need to pull some new XFS patches from the SGI guys and see if that works. The SGI kernel doesn't have this problem. Nor does any of the 9.0 kernels that mandrake has produced. Also, the cooker kernels seem to have the same problem. Thanks a lot for the additional info! It probably wouldn't have dawned on my that the only system I've built with mdk 9.1 on which the debugger works is the one that uses ext3. The other 3 use xfs! Why not simply get a GNU or Redhat kernel? Why not just buy a RedHat or Suse distribution? Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
It would seem from the thread that the problem is xfs ... MDK didn't ... problem? Arguing kernels when the problem is xfs seems hardly productive. But that's just me. James Not to be peaveish (seriously) but I was just trying to argue potential solutions from the user's viewpoint. Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
On Fri, 2003-07-04 at 13:41, Jim C wrote: It would seem from the thread that the problem is xfs ... MDK didn't ... problem? Arguing kernels when the problem is xfs seems hardly productive. But that's just me. James Not to be peaveish (seriously) but I was just trying to argue potential solutions from the user's viewpoint. And I didn't mean or desire to trash any one person. If it seemed that I did I apologize. James Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
As I cut and pasted main.c from this e-mail it wouldn't compile... do
to a number | (pipes) showing up. What I got was
i get no |'s when cut/paste... using mozilla mail. The main.c file has
no |'s in it.
#include stdlib.h
#include stdio.h
int main ( )
{
| | | | printf(hello world\n);
}
(I had to edit this to make them visible in the e-mail.)
but when I removed them.
#include stdlib.h
#include stdio.h
int main ( )
{
printf(hello world\n);
}
That is correct.
it worked fine here on 9.1 with gcc3. It's possible I guess that your
e-mail client inserted the pipes or mine did when cutting or pasting...
not sure. The important part is that it compiled, and ran, as expected.
What I'm running.
Mandrake 9.1
Kernel 2.4.21-18mdk (stock)
libgcc1-3.2.2-3mdk
gcc-cpp-3.2.2-3mdk
gcc-3.2.2-3mdk
gcc-c++-3.2.2-3mdk
[EMAIL PROTECTED] ~]$ rpm -qa | grep gcc
gcc-3.2.2-3mdk
gcc-g77-3.2.2-3mdk
gcc2.96-cpp-2.96-0.82mdk
gcc-objc-3.2.2-3mdk
gcc2.96-c++-2.96-0.82mdk
gcc-colorgcc-3.2.2-3mdk
gcc-cpp-3.2.2-3mdk
gcc2.96-2.96-0.82mdk
gcc-doc-3.2.2-3mdk
gcc-java-3.2.2-3mdk
libgcc1-3.2.2-3mdk
gcc-c++-3.2.2-3mdk
my uname -a
Linux jamlin.linuxpda.biz 2.4.21-0.18mdk #1 Wed Jun 4 12:38:37 MDT 2003
i686 unknown unknown GNU/Linux
If you note I'm not using the enterprise version of the kernel. I'm
instead using the standard. Since in my case our workstations don't
seem to benefit from it at all. (ie no performance change, or other
gain)
I use the enterprise version of the kernel because I have over 900MB of
ram. some machines have 2GB or more.
Machines running the regular kernel (smaller amount of ram) have the
same problem.
[EMAIL PROTECTED] ~]$ uname -a
Linux mirror.jpl.nasa.gov 2.4.21-0.18mdksmp #1 SMP Wed Jun 4 09:48:52
MDT 2003 i686 unknown unknown GNU/Linux
[EMAIL PROTECTED] ~]$ cat main.c
#include stdlib.h
#include stdio.h
int main ( )
{
printf(hello world\n);
}
[EMAIL PROTECTED] ~]$ gcc -g -o main main.c
[EMAIL PROTECTED] ~]$ free
total used free sharedbuffers cached
Mem:773984 766368 7616 0 52 412204
-/+ buffers/cache: 354112 419872
Swap: 20971364002096736
[EMAIL PROTECTED] ~]$ gdb ./main
GNU gdb 5.3-22mdk (Mandrake Linux)
Copyright 2002 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain
conditions.
Type show copying to see the conditions.
There is absolutely no warranty for GDB. Type show warranty for details.
This GDB was configured as i586-mandrake-linux-gnu...
(gdb) break 6
Breakpoint 1 at 0x804834c: file main.c, line 6.
(gdb) run
Starting program: /home/driver/main
Warning:
Cannot insert breakpoint -2.
Error accessing memory address 0x4000af50: Input/output error.
The same program may be running in another process.
(gdb)
When you say it worked fine what do you mean? Yes, the program runs
just fine. but I cannot trace the program in gdb, set breakpoints, or
step thru the program
I've also tried the UNI-proccess kernels and have the same problem.
None of the machines are upgraded machines, mandrake was installed
fresh via kickstart.
--
Bryan Whitehead
SysAdmin - JPL - Interferometry and Large Optical Systems
Phone: 818 354 2903
[EMAIL PROTECTED]
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
Re: [expert] kernel bug still not fixed
On Fri, 2003-06-13 at 14:40, Bryan Whitehead wrote:
Mandrake 9.1 kernels (including the latest) still have problems when
trying to use gdb. If I use any kernel from mdk9.0 the problem goes away.
[EMAIL PROTECTED] ~]$ cat main.c
#include stdlib.h
#include stdio.h
int main ( )
{
printf(hello world\n);
}
[EMAIL PROTECTED] ~]$ gcc -g main.c -o main
[EMAIL PROTECTED] ~]$ gdb main
GNU gdb 5.3-22mdk (Mandrake Linux)
Copyright 2002 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain
conditions.
Type show copying to see the conditions.
There is absolutely no warranty for GDB. Type show warranty for details.
This GDB was configured as i586-mandrake-linux-gnu...
(gdb) break 6
Breakpoint 1 at 0x804834c: file main.c, line 6.
(gdb) run
Starting program: /home/driver/main
Warning:
Cannot insert breakpoint -2.
Error accessing memory address 0x4000af50: Input/output error.
The same program may be running in another process.
(gdb)
[EMAIL PROTECTED] ~]$ uname -a
Linux mulan.jpl.nasa.gov 2.4.21-0.18mdkenterprise #1 SMP Wed Jun 4
11:44:12 MDT
2003 i686 unknown unknown GNU/Linux
[EMAIL PROTECTED] ~]$
Since our linux workstations are used heavily for development of
internal software this creates serious problems for maintaining our
workstations. We noe have to hack around getting older kernels installed
so the workstation isn't useless. Please fix this problem.
If Mandrake would like to witness this problem for themselves, I can
arrange for a visit to the lab in pasadena, ca.
Brian,
As I cut and pasted main.c from this e-mail it wouldn't compile... do
to a number | (pipes) showing up. What I got was
#include stdlib.h
#include stdio.h
int main ( )
{
| | | | printf(hello world\n);
}
(I had to edit this to make them visible in the e-mail.)
but when I removed them.
#include stdlib.h
#include stdio.h
int main ( )
{
printf(hello world\n);
}
it worked fine here on 9.1 with gcc3. It's possible I guess that your
e-mail client inserted the pipes or mine did when cutting or pasting...
not sure. The important part is that it compiled, and ran, as expected.
What I'm running.
Mandrake 9.1
Kernel 2.4.21-18mdk (stock)
libgcc1-3.2.2-3mdk
gcc-cpp-3.2.2-3mdk
gcc-3.2.2-3mdk
gcc-c++-3.2.2-3mdk
my uname -a
Linux jamlin.linuxpda.biz 2.4.21-0.18mdk #1 Wed Jun 4 12:38:37 MDT 2003
i686 unknown unknown GNU/Linux
If you note I'm not using the enterprise version of the kernel. I'm
instead using the standard. Since in my case our workstations don't
seem to benefit from it at all. (ie no performance change, or other
gain)
James
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
