When attempting to use cvs in working directories that reside on NFS4 mounted volumes, cvs fails in getcwd. If the same directory is remounted as NFS v3, cvs works. With NFS4: > # mount -t nfs4 nas:/Projects/yon /home/yon > $ cd /home/yon > $ df -h . > Filesystem Size Used Avail Use% Mounted on > nas:/Projects/yon 931G 868G 63G 94% /home/yon > $ cvs -n up VERSION > cvs [update aborted]: cannot get working directory: No such file or directory > $ cd / With NFS v3: > # umount /home/yon > # mount -t nfs nas:/exported/Projects/yon /home/yon > $ cd /home/yon > $ df -h . > Filesystem Size Used Avail Use% Mounted on > nas:/exported/Projects/yon 931G 868G 63G 94% /home/yon > $ cvs -n up VERSION > M VERSION
Created attachment 208057 [details] Other test case - logfile of "strace -f make -s". Look for pid 14647. The problem seems to be not gentoo-specific. I am experiencing the same issues under Fedora 11 as well. For reference, please see https://bugzilla.redhat.com/show_bug.cgi?id=530625
Debian bug: http://bugs.debian.org/459641
(In reply to comment #2) > Debian bug: > http://bugs.debian.org/459641 Isn't it the same problem as http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=461154 ? I had the same issue when using cvs in chroot where is directory mounted with mount --bind. It's my directory for cross compilation and I need cvs for autopoint. It was resolved in masked cvs-1.12.13.1. But because of new glibc (I have 2.11-r1) I had to use extra patch solving '*** %n in writable segment detected ***' error, report found here: https://qa.mandriva.com/show_bug.cgi?id=31848 Patch found http://www.mail-archive.com/pld-cvs-commit@lists.pld-linux.org/msg173504.html I'll attach reformated patch to this bug which can be easily used from ebuild with: epatch "${FILESDIR}"/${PN}-1.12.13-printf.patch Regards,
Created attachment 219971 [details, diff] cvs-1.12.13-printf.patch Patch for segfaulting cvs with '*** %n in writable segment detected ***' error message.
I see the summary was changed from dev-util/cvs... into dev-vcs/cvs... - however I cannot see any comments or explanations for this change - what is the relation of /dev/vcs (i. e. virtual console terminals) to this problem?
(In reply to comment #5) > I see the summary was changed from dev-util/cvs... into dev-vcs/cvs... - > however I cannot see any comments or explanations for this change - what is the > relation of /dev/vcs (i. e. virtual console terminals) to this problem? all cvs ebuilds were moved from dev-util category to dev-vcs. So someone is cleaning bugzilla to show correct category. FYI: vcs is used here as "version control system" not virtual console..
I'm experiencing what I believe to be the same issue on a CIFS share. Error returned from CVS is: "cvs [update aborted]: cannot get working directory: No such file or directory" Upgrading to masked version (1.12.13.1) seems to fix it.
Actually, now I'm getting the "*** %n in writable segment detected ***" error, so still not usable.
1.12.13.1 with the attached patch works for me so far.
Yet again I spoke too soon. I've now seeing the following error upon cvs update: 556 12:03 $ cvs update cvs [update aborted]: could not chdir to /shared/design/cvs_src/dev_utils: No such file or directory strace reveals: close(5) = 0 chdir("/shared/design/jbramley/cvs_src/MASTERPROJECT/Ladbrokes/Madrid") = -1 ENOENT (No such file or directory) write(2, "cvs [update aborted]: could not "..., 131cvs [update aborted]: could not chdir to /shared/design/jbramley/cvs_src/MASTERPROJECT/Ladbrokes/Madrid: No such file or directory ) = 131 rt_sigprocmask(SIG_SETMASK, ~[RTMIN RT_1], [], 8) = 0 close(1) = 0 exit_group(1) = ?
please retry with 1.12.12-r12