Eric Abrahamsen <e...@ericabrahamsen.net> writes: > Daimrod <daim...@gmail.com> writes: > >> Bastien <b...@gnu.org> writes: >> >>> Hi Eric, >>> >>> Eric Abrahamsen <e...@ericabrahamsen.net> writes: >>> >>>> After Nicolas made the last round of improvements to the caching >>>> mechanism I got far fewer hangs with Org, but they are still happening. >>>> Maybe once a day or so, on average, editing something in an Org buffer >>>> causes emacs to hang, and my fans to spin up, and there we are until I >>>> kill emacs. > > [...] > >> By the way, if you want to see in which part the infloop occurs, you can >> attach a gdb debugger to the running emacs, source the >> <path-to-emacs-source>/src/.gdbinit file and use the `xbacktrace' command. >> >> $ gdb <path-to-emacs-executable> <emacs-pid> >> gdb) source <path-to-emacs-source>/src/.gdbinit >> ... >> gdb) xbacktrace >> >> You can also use the `bt' command but it contains much more noise. > > I got another one just now (while moving from one org table cell to the > next), and that was the gdb backtrace: > > "avl-tree--do-delete" (0xbfffe858) > "avl-tree-delete" (0xbfffe998) > "byte-code" (0xbfffeaa0) > "byte-code" (0xbfffec30) > "org-element--cache-process-request" (0xbfffedd8) > "byte-code" (0xbfffeef0) > "org-element--cache-sync" (0xbffff0a8) > "org-element-at-point" (0xbffff1e8) > "org-mode-flyspell-verify" (0xbffff338) > "flyspell-word" (0xbffff478) > "byte-code" (0xbffff580) > "flyspell-post-command-hook" (0xbffff784)
It seems the lockup also happens in `org-element--cache-...'. > Not much, and probably not that useful. I'll start running org > uncompiled, and try the debug-on-event trick. Thanks for you time! > FWIW, this was the first lockup that *didn't* occur in a logbook > drawer -- that's where I usually get them. Either a full lockup, or else > the cache goes wonky so that adding log notes (or even just navigating > in the drawer) gives me that "bound on wrong side of point" you get when > you try to search forwards, backwards. That's weird (in my cases it usually mess up with the input-method). -- Daimrod/Greg