Bug#325600: Patch for buggy pthreads on stable found
Uwe Schindler a écrit : > Hi again, > >> I forgot to mention: >> >> Someone could try to enable the test suite on package build again. Now the >> linuxthreads testcases should pass... >> >> This can be enabled by changing debian/sysdeps/alpha.mk ! >> >> I will not do this as the package build takes too long on my machine and I >> do not want wait 3.5 hours again and then 5 hours for the tests to >> complete. > > The tests seem to run now, no more failures in (I haven't tested this > extensive, I only ran "make -C build-tree/alpha-libc/ -k check"): > > linuxthreads/ex7,ex9,ex10,ex18 > > Because of that this can be also removed from debian/sysdeps/alpha.mk: > > # disabled because alpha linuxthreads ex7, ex9, ex10, ex18 stopps eternally > # even if TIMEOUTSCALE is set - SIGALRM from the program is just ignored. > RUN_TESTSUITE = no > Thanks for the test. Unfortunately, reenabling the testsuite in a stable upload gives too much risk of a failed build. Therefore I prefer to leave it as now. Note that test testsuite is enabled again in unstable (NPTL version). Bye, Aurelien -- .''`. Aurelien Jarno | GPG: 1024D/F1BCDB73 : :' : Debian developer | Electrical Engineer `. `' [EMAIL PROTECTED] | [EMAIL PROTECTED] `-people.debian.org/~aurel32 | www.aurel32.net -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#325600: Patch for buggy pthreads on stable found
Uwe Schindler a écrit : > Tags: patch > > I created a patch for glibc (version 2.3.6.ds1) that fixes the zombie > processes bug. It uses an ifdef'ed version of the waitpid_not_cancel macro > that uses on alpha the standard wait4 syscall instead of an inline syscall. > All other architectures are not changed. > > I put the attached patch into debian/patches/alpha/ of the source package > and added the following line to debian/patches/series: > > alpha/local-waitpid-not-cancel.diff -p0 > > after that I built with "dpkg-buildpackage -rfakeroot". The .deb files and > source packages can be found here: > Thanks for the patch, I have checked in our SVN repository for the etch branch. If the stable release managers agree, this patch will be included in etch r1. Bye, Aurelien -- .''`. Aurelien Jarno | GPG: 1024D/F1BCDB73 : :' : Debian developer | Electrical Engineer `. `' [EMAIL PROTECTED] | [EMAIL PROTECTED] `-people.debian.org/~aurel32 | www.aurel32.net -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#325600: Patch for buggy pthreads on stable found
Hi again, > I forgot to mention: > > Someone could try to enable the test suite on package build again. Now the > linuxthreads testcases should pass... > > This can be enabled by changing debian/sysdeps/alpha.mk ! > > I will not do this as the package build takes too long on my machine and I > do not want wait 3.5 hours again and then 5 hours for the tests to > complete. The tests seem to run now, no more failures in (I haven't tested this extensive, I only ran "make -C build-tree/alpha-libc/ -k check"): linuxthreads/ex7,ex9,ex10,ex18 Because of that this can be also removed from debian/sysdeps/alpha.mk: # disabled because alpha linuxthreads ex7, ex9, ex10, ex18 stopps eternally # even if TIMEOUTSCALE is set - SIGALRM from the program is just ignored. RUN_TESTSUITE = no -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#325600: Patch for buggy pthreads on stable found
Tags: patch I created a patch for glibc (version 2.3.6.ds1) that fixes the zombie processes bug. It uses an ifdef'ed version of the waitpid_not_cancel macro that uses on alpha the standard wait4 syscall instead of an inline syscall. All other architectures are not changed. I put the attached patch into debian/patches/alpha/ of the source package and added the following line to debian/patches/series: alpha/local-waitpid-not-cancel.diff -p0 after that I built with "dpkg-buildpackage -rfakeroot". The .deb files and source packages can be found here: http://alpha.thetaphi.de/glibc-alpha/ Anybody who uses Debian stable and has the zombiefied processes should install the file http://alpha.thetaphi.de/glibc-alpha/libc6.1_2.3.6.ds1-13_alpha.deb On their machine to get rid of the error. The other deb files are the same like from the repository (so you do only need to install libc6.1). The version number of the package is the same (I did not change the package description). After installation I started some known-to-zombiefy programs (gnome-cups-icon, amule, mysql) and no further zombies occurred. To check if the patched version of glibc was really used, I looked into ps output and checked that multithreaded programs create more than one process entry (as not nptl) but do not leave them zombiefied when cleaning up. As this patch seems to fix this bug I stop now further investigating and let Aurelien and the Maintainers add the patch! Hope I helped you! My machine will now go again to 2.5-2 with NPTL :) ... - Uwe Schindler H.-H.-Meier-Allee 63, D-28213 Bremen http://www.thetaphi.de eMail: [EMAIL PROTECTED] local-waitpid-not-cancel.diff Description: Binary data
Bug#325600: Patch for buggy pthreads on stable found
I forgot to mention: Someone could try to enable the test suite on package build again. Now the linuxthreads testcases should pass... This can be enabled by changing debian/sysdeps/alpha.mk ! I will not do this as the package build takes too long on my machine and I do not want wait 3.5 hours again and then 5 hours for the tests to complete. I found no way to only run the tests of linuxthreads alone (how can this be done with manually starting make without building the package again???) -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
