At our site, on many, but not all Fedora 7 boxes, launching emacs (gui, not text-only console) crashes semi-regularly (~50%) on startup. For example, my konsole log now contains: [rdieter1@sting ~]$ emacs Fatal error (11)Segmentation fault [rdieter1@sting ~]$ emacs Fatal error (11)Segmentation fault [rdieter1@sting ~]$ emacs Fatal error (11)Segmentation fault [rdieter1@sting ~]$ emacs Fatal error (4)Illegal instruction [rdieter1@sting ~]$ emacs Fatal error (11)Segmentation fault [rdieter1@sting ~]$ emacs Fatal error (4)Illegal instruction [rdieter1@sting ~]$ emacs Fatal error (11)Segmentation fault [rdieter1@sting ~]$ emacs (yay, it worked!) I've tried using emacs-22.0.990-2.fc7 and emacs-22.1-1.fc7.i386 in updates-testing. They both seem to exhibit the problem equally.
Three questions: 1. What version of glibc is installed on these systems? 2. Does it happen if you run "setarch i386 -R emacs"? 3. Can you supply a core? Chip
1. box 1 $ rpm -q emacs glibc emacs-22.1-1.fc7.i386 glibc-2.6-3.i686 box 2 $rpm -q emacs glibc emacs-22.0.990-2.fc7.i386 glibc-2.6-3.i686 2. yes, just tried that. [rdieter1@sting ~]$ setarch i386 -R emacs (emacs:7704): -CRITICAL **: \u0004: assertion `\xff\xff\xff\xff' failed Fatal error (11)Segmentation fault $ setarch i386 -R emacs Fatal error (11)Segmentation fault (then it worked once...) 3. sure, it's big tho, almost 2MB gzipped, attach here? (is emacs-debuginfo required for that?)
Created attachment 158969 [details] emacs SIGSEGV core
fyi, core sampled on box with updates-testing enabled w/emacs-22.1-1.fc7
(In reply to comment #3) > Created an attachment (id=158969) [edit] > emacs SIGSEGV core > One more question: did you prelink this binary? Chip
These systems have prelinking enabled, yes. I also tried yum remove emacs yum install emacs (to hopefully undo any prelinking), but that didn't seem to help any.
(In reply to comment #6) > These systems have prelinking enabled, yes. Don't do that. See bug 227354. > I also tried > yum remove emacs > yum install emacs > (to hopefully undo any prelinking), but that didn't seem to help any. Try /usr/sbin/prelink -ua instead. Chip
Ouch, turn off prelinking universally just to use emacs? :)
(In reply to comment #8) > Ouch, turn off prelinking universally just to use emacs? :) No, turn off prelinking universally to verify that is the problem you are having with emacs. Then if that fixes it, edit /etc/prelink.conf to add a line containing -b /usr/bin/emacs-*
Ah, duh, sorry, will do. (I'm betting it'll do the trick).
After 'prelink -ua', emacs now starts reliably. Adding -b /usr/bin/emacs-* and re-prelinking to see how things go.
Crud, re-prelinking, even with -b /usr/bin/emacs-*,still yielded an unstable emacs.
(In reply to comment #12) > Crud, re-prelinking, even with -b /usr/bin/emacs-*,still yielded an unstable emacs. I notice that prelinking emacs seems to work fine on RHEL-5. So something about prelinking is breaking emacs; here we go back into the dumper again. Chip
Haven't seen any segfaults in awhile, so I'll close this (for now).