Reproducible random python crash.
Kyle, Helge, John, James, Could I get each of you to follow this recipe and tell me if python crashes or hangs randomly during the execution of the last step? Recipe: mkdir -p ~/python-test cd ~/python-test apt-get update apt-get build-dep xmms2 apt-get source xmms2 cd xmms2_0.6DrMattDestruction ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac Running the last step on a 2.6.26-1-parisc64-smp kernel results in random behaviour ranging from sigsegv, sigill, sigtrap, and a hang. Given that each of you might run it on a different kernel, I was hoping to see if the behaviour is the same across different kernels. Thanks! Cheers, Carlos. -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
On Fri, Aug 21, 2009 at 10:59 AM, Carlos O'Donell wrote: > Could I get each of you to follow this recipe and tell me if python > crashes or hangs randomly during the execution of the last step? > > Recipe: > mkdir -p ~/python-test > cd ~/python-test > apt-get update > apt-get build-dep xmms2 > apt-get source xmms2 > cd xmms2_0.6DrMattDestruction > ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man > --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl > --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux > --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python > --without-plugins=mac > > Running the last step on a 2.6.26-1-parisc64-smp kernel results in > random behaviour ranging from sigsegv, sigill, sigtrap, and a hang. > Given that each of you might run it on a different kernel, I was > hoping to see if the behaviour is the same across different kernels. On 2.6.31-rc2-00010-g22a5b0c-dirty (kyle's 4 CPU PA8800) I get a hang and segfaults, the segfault is reproducible the hang happens every once in a while. Cheers, Carlos. -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
On Fri, 2009-08-21 at 11:43 -0400, Carlos O'Donell wrote: > On Fri, Aug 21, 2009 at 10:59 AM, Carlos > O'Donell wrote: > > Could I get each of you to follow this recipe and tell me if python > > crashes or hangs randomly during the execution of the last step? > > > > Recipe: > > mkdir -p ~/python-test > > cd ~/python-test > > apt-get update > > apt-get build-dep xmms2 > > apt-get source xmms2 > > cd xmms2_0.6DrMattDestruction > > ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man > > --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl > > --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux > > --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python > > --without-plugins=mac > > > > Running the last step on a 2.6.26-1-parisc64-smp kernel results in > > random behaviour ranging from sigsegv, sigill, sigtrap, and a hang. > > Given that each of you might run it on a different kernel, I was > > hoping to see if the behaviour is the same across different kernels. > > On 2.6.31-rc2-00010-g22a5b0c-dirty (kyle's 4 CPU PA8800) I get a hang > and segfaults, the segfault is reproducible the hang happens every > once in a while. Same here on ion (2.6.31-rc6). (no hang seen, just segfaults in python) James -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
> > ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man > > --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl > > --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux > > --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python > > --without-plugins=mac > > > > Running the last step on a 2.6.26-1-parisc64-smp kernel results in > > random behaviour ranging from sigsegv, sigill, sigtrap, and a hang. > > Given that each of you might run it on a different kernel, I was > > hoping to see if the behaviour is the same across different kernels. > > On 2.6.31-rc2-00010-g22a5b0c-dirty (kyle's 4 CPU PA8800) I get a hang > and segfaults, the segfault is reproducible the hang happens every > once in a while. Do you have Grant's last patch installed? Although it's early yet to say for sure, I think it may have fixed the random segv problem on gsyprf11 with 2.6.30.5. Dave -- J. David Anglin dave.ang...@nrc-cnrc.gc.ca National Research Council of Canada (613) 990-0752 (FAX: 952-6602) -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
Carlos O'Donell wrote: > ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man > --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl > --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux > --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python > --without-plugins=mac I've given the command 2 tries with each kernel. 64-bit SMP kernels == Debian 2.6.24-1-parisc64-smp Checking for libsidplay2 : ok Fatal Python error: PyEval_RestoreThread: NULL tstate Checking for library sidplay2 : ../test: line 7: 1870 Aborted ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac Checking for libmpg123 >= 1.5.1 : ok Testing mpg123 with default off_t : failed Fatal Python error: PyEval_RestoreThread: NULL tstate Testing mpg123 with large off_t : ../test: line 7: 2021 Aborted ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac Upstream 2.6.30-rc5 --- Checking for header sys/soundcard.h : ok ./test: line 7: 25852 Trace/breakpoint trap ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac Checking for vorbisfile : ok ./test: line 7: 25980 Segmentation fault ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac Upstream 2.6.31-rc6 (current git) --- Checking for libpulse: ok ../test: line 7: 1974 Segmentation fault ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac Checking for vorbisenc: ok ../test: line 7: 2156 Segmentation fault ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac 32-bit SMP kernels == Debian 2.6.26-2-parisc-smp -- Checking for header sys/soundcard.h : ok ../test: line 7: 2040 Segmentation fault ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac Checking for header sys/soundcard.h : ok ../test: line 7: 2074 Illegal instruction ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python --without-plugins=mac Debian 2.6.30-1-parisc-smp -- fails to boot 64-bit NON-SMP kernel = Debian 2.6.30-1-parisc64 Command runs fine. I'm going to start a build of xmms2 on my box with this kernel and (if successful) will do a porter upload so that the transition to testing can proceed. John David Anglin wrote: > Do you have Grant's last patch installed? Although it's early yet > to say for sure, I think it may have fixed the random segv problem > on gsyprf11 with 2.6.30.5. What patch is that? Link? Cheers, FJP -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
On Fri, Aug 21, 2009 at 12:19 PM, John David Anglin wrote: >> > ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man >> > --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl >> > --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux >> > --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python >> > --without-plugins=mac >> > >> > Running the last step on a 2.6.26-1-parisc64-smp kernel results in >> > random behaviour ranging from sigsegv, sigill, sigtrap, and a hang. >> > Given that each of you might run it on a different kernel, I was >> > hoping to see if the behaviour is the same across different kernels. >> >> On 2.6.31-rc2-00010-g22a5b0c-dirty (kyle's 4 CPU PA8800) I get a hang >> and segfaults, the segfault is reproducible the hang happens every >> once in a while. > > Do you have Grant's last patch installed? Although it's early yet > to say for sure, I think it may have fixed the random segv problem > on gsyprf11 with 2.6.30.5. Which patch is this? Also note that I ran this with LD_BIND_NOW=1 to force the dynamic loader to resolve all relocations before running the program, and it still crashes. Cheers, Carlos. -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
> >> On 2.6.31-rc2-00010-g22a5b0c-dirty (kyle's 4 CPU PA8800) I get a hang > >> and segfaults, the segfault is reproducible the hang happens every > >> once in a while. > > > > Do you have Grant's last patch installed? =A0Although it's early yet > > to say for sure, I think it may have fixed the random segv problem > > on gsyprf11 with 2.6.30.5. > > Which patch is this? http://article.gmane.org/gmane.linux.ports.parisc/2047 Since I wrote, I had one segv with 2.6.30.5 in the GCC libjava testsuite. However, I think things are improved. I have managed to get through most of a GCC build and check with a SMP kernel. I tried to follow your recipe on gsyprf11, but the source for 0.5 was dounloaded. I didn't have any trouble with the configure command with this version. Dave -- J. David Anglin dave.ang...@nrc-cnrc.gc.ca National Research Council of Canada (613) 990-0752 (FAX: 952-6602) -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
On Fri, Aug 21, 2009 at 3:30 PM, John David Anglin wrote: >> >> On 2.6.31-rc2-00010-g22a5b0c-dirty (kyle's 4 CPU PA8800) I get a hang >> >> and segfaults, the segfault is reproducible the hang happens every >> >> once in a while. >> > >> > Do you have Grant's last patch installed? =A0Although it's early yet >> > to say for sure, I think it may have fixed the random segv problem >> > on gsyprf11 with 2.6.30.5. >> >> Which patch is this? > > http://article.gmane.org/gmane.linux.ports.parisc/2047 > > Since I wrote, I had one segv with 2.6.30.5 in the GCC libjava > testsuite. However, I think things are improved. I have managed to > get through most of a GCC build and check with a SMP kernel. > > I tried to follow your recipe on gsyprf11, but the source for 0.5 > was dounloaded. I didn't have any trouble with the configure command > with this version. I should have mentioned you need unstable sources (not unstable debs) in your sources.list. e.g. deb-src ftp://ftp.ca.debian.org/debian/ unstable main contrib non-free Does that help? Cheers, Carlos. -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
> I should have mentioned you need unstable sources (not unstable debs) > in your sources.list. > e.g. deb-src ftp://ftp.ca.debian.org/debian/ unstable main contrib non-free > Does that help? Yes. segvs and hangs on gsyprf11. Think the hangs occur when rerunning configure. I looked at one segv and it appeared r19 was corrupt causing a branch to never-never-land. Dave -- J. David Anglin dave.ang...@nrc-cnrc.gc.ca National Research Council of Canada (613) 990-0752 (FAX: 952-6602) -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
> Kyle, Helge, John, James, > > Could I get each of you to follow this recipe and tell me if python > crashes or hangs randomly during the execution of the last step? > > Recipe: > mkdir -p ~/python-test > cd ~/python-test > apt-get update > apt-get build-dep xmms2 > apt-get source xmms2 > cd xmms2_0.6DrMattDestruction > ./waf configure --nocache --prefix=/usr --with-mandir=/usr/share/man > --with-perl-archdir=/usr/lib/perl5 --with-perl-binary=/usr/bin/perl > --with-ruby-archdir=/usr/lib/ruby/1.8/hppa-linux > --with-ruby-libdir=/usr/lib/ruby/1.8 --without-optionals=python > --without-plugins=mac > > Running the last step on a 2.6.26-1-parisc64-smp kernel results in > random behaviour ranging from sigsegv, sigill, sigtrap, and a hang. > Given that each of you might run it on a different kernel, I was > hoping to see if the behaviour is the same across different kernels. I ran the apt-get though configure commands several times with no failures on "Linux mx3210 2.6.30.5 #7 Wed Aug 19 17:31:54 EDT 2009 parisc64". So, it looks like the problem is SMP specific. Same kernel with SMP config on gsyprf11 resulted in random segs and hangs. The hangs may depend on the presence of cached data used by configure. I booted 2.6.30.5 from gsyprf11 because I had one random segv and one command hang while running the GCC testsuite. I suspect that some hangs are due to dropped signals. I'm tempted to try the sequence with 2.6.22.19. Dave -- J. David Anglin dave.ang...@nrc-cnrc.gc.ca National Research Council of Canada (613) 990-0752 (FAX: 952-6602) -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
> I booted 2.6.30.5 from gsyprf11 because I had one random segv and > one command hang while running the GCC testsuite. I suspect that some > hangs are due to dropped signals. I'm tempted to try the sequence > with 2.6.22.19. SMP 2.6.22.19 also fails with random segvs and hangs. Dave -- J. David Anglin dave.ang...@nrc-cnrc.gc.ca National Research Council of Canada (613) 990-0752 (FAX: 952-6602) -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
On Sun, Aug 23, 2009 at 5:21 PM, John David Anglin wrote: >> I booted 2.6.30.5 from gsyprf11 because I had one random segv and >> one command hang while running the GCC testsuite. I suspect that some >> hangs are due to dropped signals. I'm tempted to try the sequence >> with 2.6.22.19. > > SMP 2.6.22.19 also fails with random segvs and hangs. Thanks Dave. Any idea what might be the cause? I'm still setting up my system so I can debug this. Cheers, Carlos. -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
> Any idea what might be the cause? No. I looked at two core dumps briefly. In the first, it looked like the PIC register value was corrupt causing a transfer to page 0. In the second, there wasn't any backtrace info (both pc and rp were 0). Unfortunately, this doesn't say much. It would be fantastic if you could narrow this down. Dave -- J. David Anglin dave.ang...@nrc-cnrc.gc.ca National Research Council of Canada (613) 990-0752 (FAX: 952-6602) -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
On Fri, Aug 21, 2009 at 07:36:47PM +0200, Frans Pop wrote: ... > John David Anglin wrote: > > Do you have Grant's last patch installed? Although it's early yet > > to say for sure, I think it may have fixed the random segv problem > > on gsyprf11 with 2.6.30.5. > > What patch is that? Link? here --> http://article.gmane.org/gmane.linux.ports.parisc/2047 But appended below again for convenience. Apologies for the late reply. B0rkage in debian grub2 update process rendered my x86 laptop unbootable. I've ended up re-installing the root partition and slowly restoring misc config (sound/video/firewall/etc) and ignoring email until that was done. thanks, grant diff --git a/arch/parisc/kernel/traps.c b/arch/parisc/kernel/traps.c index 528f0ff..8b58bf0 100644 --- a/arch/parisc/kernel/traps.c +++ b/arch/parisc/kernel/traps.c @@ -532,7 +532,7 @@ void notrace handle_interruption(int code, struct pt_regs *regs) /* Kill the user process later */ regs->iaoq[0] = 0 | 3; regs->iaoq[1] = regs->iaoq[0] + 4; - regs->iasq[0] = regs->iasq[0] = regs->sr[7]; + regs->iasq[0] = regs->iasq[1] = regs->sr[7]; regs->gr[0] &= ~PSW_B; return; } -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
Re: Reproducible random python crash.
On Friday 28 August 2009, Grant Grundler wrote: > On Fri, Aug 21, 2009 at 07:36:47PM +0200, Frans Pop wrote: > > John David Anglin wrote: > > > Do you have Grant's last patch installed? Although it's early yet > > > to say for sure, I think it may have fixed the random segv problem > > > on gsyprf11 with 2.6.30.5. > > > > What patch is that? Link? > > here --> http://article.gmane.org/gmane.linux.ports.parisc/2047 Thanks. Somebody else pointed it out as well. I gave it a try, but it did not make any difference for me for these python errors. Cheers, FJP -- To UNSUBSCRIBE, email to debian-hppa-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org