Re: GC bug?

[Stephen Berman]

On Sat, 22 Oct 2005 11:51:30 -0400 "Richard M. Stallman" <address@hidden> wrote:
[Stephen Berman wrote:]

>     I made a diary entry and then called calendar and Emacs immediately
>     hung or went into an infinite loop, presumably in diary-list-entries,
>     since the last message was "Preparing diary...".  I could get out of
>     the loop by typing C-g.

It has happened again (in a different Emacs session, I had to close the
one where it happened before).  Here are the particular circumstances,
FWIW: I have a file which is an #include in my ~/diary.  Currently
this file is changed but not saved.  When I type `d'
(diary-view-entries) in Calendar on the date which is the date of the
last entry in the unsaved included file, Emacs goes into an infinite
loop, as described above.  If I type `d' on any other date in
Calendar, there is no problem.  Also, if I save the included file and
then type `d' on the date of its last entry, there is no problem.

> Then it probably isn't a GC bug.  It is probably an infinite loop
> that does a lot of consing and therefore frequently GCs.
>
> I suggest you stop it outside GC with a breakpoint at Fgarbage_collect
> and then type `finish' to get over the GC.  Then please follow the
> methods suggested in etc/DEBUG for debugging infinite loops.

I tried to do this.  Fgarbage_collect is called at several places in
the Emacs C sources, but I assume you mean the invocation in alloc.c,
so that's where I put the breakpoint.  When I set the breakpoint
before inducing the infinite loop, then gdb hits the breakpoint
immediately, so I cannot induce loop.  So I first induced the loop and
then set the breakpoint, and this is the result:
(gdb) c
Continuing.

[1]+  Stopped                 gdb emacs 16198
address@hidden:~> fg
gdb emacs 16198

Program received signal SIGINT, Interrupt.
[Switching to Thread 1084223872 (LWP 16198)]
Flooking_at (regexp=144421059) at search.c:354
354       return looking_at_1 (regexp, 0);
(gdb) break /home/steve/cvsroot/emacs/src/alloc.c:Fgarbage_collect
Breakpoint 1 at 0x813d4fc: file alloc.c, line 4695.
(gdb) finish
Run till exit from #0  Flooking_at (regexp=144421059) at search.c:354
Ffuncall (nargs=2, args=0xbfe57660) at eval.c:2870
2870              goto done;
Value returned is $1 = 137379905
(gdb) finish
Run till exit from #0  Ffuncall (nargs=2, args=0xbfe57660) at eval.c:2870
Fbyte_code (bytestr=144421571, vector=144365692, maxdepth=9) at bytecode.c:692
692                 break;
Value returned is $2 = 137379905
(gdb) finish
Run till exit from #0  Fbyte_code (bytestr=144421571, vector=144365692,
    maxdepth=9) at bytecode.c:692

At this point Emacs started consuming all available CPU cycles and I
had to interrupt gdb.  So I don't know what else to do; can you or
someone else advise?  Further investigation will have to wait for the
infinite loop to occur again, since I have to shut down the computer.
But since it happened in two consecutive Emacs sessions, the wait may
not be too long; I will also try the recipe I described above.

Steve Berman