Bug#508914: cowdancer: cowbuilder does not separate locks in different chroots
Version: 0.47 On Mon, 23 Mar 2009 17:30:36 +0100 Stefan Fritschwrote: > On Friday 19 December 2008, Junichi Uekawa wrote: >> Ah, I think I remember something about ubuntu sudo environmental >> variable being broken. Is LD_PRELOAD properly passed through sudo? >> I think there was some configuration item you had to do. > > Sudo is not necessary to reproduce the bug. > > However upgrading to cowdancer 0.47 and pbuilder and > 0.181 improved the situation. The build no longer fails, but there is > still some error message about a debconf lock file. > > I won't have time in the near future to investigate this further. If > you want, you can close this bug as fixed in 0.47. I just tried two builds running at the same time and could not reproduce this. Regards, James signature.asc Description: Message signed with OpenPGP using GPGMail
Bug#508914: cowdancer: cowbuilder does not separate locks in different chroots
On Friday 19 December 2008, Junichi Uekawa wrote: Ah, I think I remember something about ubuntu sudo environmental variable being broken. Is LD_PRELOAD properly passed through sudo? I think there was some configuration item you had to do. Sudo is not necessary to reproduce the bug. However upgrading to cowdancer 0.47 and pbuilder and 0.181 improved the situation. The build no longer fails, but there is still some error message about a debconf lock file. I won't have time in the near future to investigate this further. If you want, you can close this bug as fixed in 0.47. Cheers, Stefan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#508914: cowdancer: cowbuilder does not separate locks in different chroots
Ah, I think I remember something about ubuntu sudo environmental variable being broken. Is LD_PRELOAD properly passed through sudo? I think there was some configuration item you had to do. At Thu, 18 Dec 2008 08:38:19 +0900, Junichi Uekawa wrote: At Wed, 17 Dec 2008 21:21:47 +0100, Stefan Fritsch wrote: On Wednesday 17 December 2008, Junichi Uekawa wrote: Package: cowdancer Version: 0.44 Why is this so old? This is actually Ubuntu hardy. I couldn't reproduce it now at home with sid / 0.48 but the setup is quite different, too. I've noticed that too. What's the lock function they use? This comes from apt/libapt-pkg: apt-pkg/contrib/fileutl.cc: int FD = open(File.c_str(),O_RDWR | O_CREAT | O_NOFOLLOW,0640); ... fl.l_type = F_WRLCK; fl.l_whence = SEEK_SET; fl.l_start = 0; fl.l_len = 0; if (fcntl(FD,F_SETLK,fl) == -1) Since it does an open O_RDWR first, maybe cowdancer tried to copy the file but fails? There was a cp related bug fix in 0.47. I didn't see any WIFEXITED related error message, though. Do you have any idea how I could verify if it is the same bug, short of installing a backported newer cowdancer? You should install backported cowdancer or reproduce with the latest version; the version in Ubuntu is not supported by anybody, and later versions have additional debug messages etc. By setting COWDANCER_DEBUG environmental variable, you get more debug output which might or might not help. Thanks for your help. Cheers, Stefan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#508914: cowdancer: cowbuilder does not separate locks in different chroots
Hi, At Tue, 16 Dec 2008 17:39:11 +0100, Stefan Fritsch wrote: Package: cowdancer Version: 0.44 Why is this so old? Severity: normal When several cowbuilder instances are started in parallel, dpkg/aptitude in each instance will try to lock its lock file. However, since the files are linked across chroots, only the first succeeds and all other fail with Could not get lock /var/lib/dpkg/lock - open (11 Resource temporarily unavailable) The strace output is fcntl64(4, F_SETLK, {type=F_WRLCK, whence=SEEK_SET, start=0, len=0}) = -1 EAGAIN (Resource temporarily unavailable) As a workaround, it seems to be possible to do rm /var/cache/pbuilder/base.cow/var/lib/dpkg/lock rm /var/cache/pbuilder/base.cow/var/lock/aptitude But it would be better if cowdancer would implicitly copy a file if a process in the chroot tries to lock it. I've noticed that too. What's the lock function they use? regards, junichi -- dan...@{netfort.gr.jp,debian.org} -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#508914: cowdancer: cowbuilder does not separate locks in different chroots
On Wednesday 17 December 2008, Junichi Uekawa wrote: Package: cowdancer Version: 0.44 Why is this so old? This is actually Ubuntu hardy. I couldn't reproduce it now at home with sid / 0.48 but the setup is quite different, too. I've noticed that too. What's the lock function they use? This comes from apt/libapt-pkg: apt-pkg/contrib/fileutl.cc: int FD = open(File.c_str(),O_RDWR | O_CREAT | O_NOFOLLOW,0640); ... fl.l_type = F_WRLCK; fl.l_whence = SEEK_SET; fl.l_start = 0; fl.l_len = 0; if (fcntl(FD,F_SETLK,fl) == -1) Since it does an open O_RDWR first, maybe cowdancer tried to copy the file but fails? There was a cp related bug fix in 0.47. I didn't see any WIFEXITED related error message, though. Do you have any idea how I could verify if it is the same bug, short of installing a backported newer cowdancer? Thanks for your help. Cheers, Stefan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#508914: cowdancer: cowbuilder does not separate locks in different chroots
At Wed, 17 Dec 2008 21:21:47 +0100, Stefan Fritsch wrote: On Wednesday 17 December 2008, Junichi Uekawa wrote: Package: cowdancer Version: 0.44 Why is this so old? This is actually Ubuntu hardy. I couldn't reproduce it now at home with sid / 0.48 but the setup is quite different, too. I've noticed that too. What's the lock function they use? This comes from apt/libapt-pkg: apt-pkg/contrib/fileutl.cc: int FD = open(File.c_str(),O_RDWR | O_CREAT | O_NOFOLLOW,0640); ... fl.l_type = F_WRLCK; fl.l_whence = SEEK_SET; fl.l_start = 0; fl.l_len = 0; if (fcntl(FD,F_SETLK,fl) == -1) Since it does an open O_RDWR first, maybe cowdancer tried to copy the file but fails? There was a cp related bug fix in 0.47. I didn't see any WIFEXITED related error message, though. Do you have any idea how I could verify if it is the same bug, short of installing a backported newer cowdancer? You should install backported cowdancer or reproduce with the latest version; the version in Ubuntu is not supported by anybody, and later versions have additional debug messages etc. By setting COWDANCER_DEBUG environmental variable, you get more debug output which might or might not help. Thanks for your help. Cheers, Stefan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#508914: cowdancer: cowbuilder does not separate locks in different chroots
Package: cowdancer Version: 0.44 Severity: normal When several cowbuilder instances are started in parallel, dpkg/aptitude in each instance will try to lock its lock file. However, since the files are linked across chroots, only the first succeeds and all other fail with Could not get lock /var/lib/dpkg/lock - open (11 Resource temporarily unavailable) The strace output is fcntl64(4, F_SETLK, {type=F_WRLCK, whence=SEEK_SET, start=0, len=0}) = -1 EAGAIN (Resource temporarily unavailable) As a workaround, it seems to be possible to do rm /var/cache/pbuilder/base.cow/var/lib/dpkg/lock rm /var/cache/pbuilder/base.cow/var/lock/aptitude But it would be better if cowdancer would implicitly copy a file if a process in the chroot tries to lock it. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org