random core dumps

random core dumps

[Bob Halley]

Red Hat GNU/Linux 8.0, Marcelo 2.4.20, on a dual Athlon.

Lately (the last week or so) I've been experiencing core dumps from
the CVS mainline emacs.  They occur randomly, and seemingly most often
when starting emacs.

All of the cores have been in

        compact_small_strings () at alloc.c:1608

At the time of the crash, the debugger reports that "from" is
(struct sdata *) 0xb.  I haven't been able to figure out why this is
happening yet.

Here's some more of the stack trace.  Let me know if there's anything
else I can do to help track this down.

Core was generated by `emacs'.
Program terminated with signal 11, Segmentation fault.
...
#0  0x42028cc1 in kill () from /lib/i686/libc.so.6
#1  0x080d46f0 in fatal_error_signal (sig=7444) at emacs.c:379
#2  <signal handler called>
#3  compact_small_strings () at alloc.c:1608
#4  0x0811cb43 in gc_sweep () at alloc.c:5005
#5  0x0811bbec in Fgarbage_collect () at alloc.c:4248
#6  0x0812ed54 in Feval (form=1482274492) at eval.c:1989
#7  0x08141d84 in readevalloop (readcharfun=405588356, stream=0x858fd18, sourcename=945344908, 
    evalfun=0x812e8a0 <Feval>, printflag=0, unibyte=0, readfun=405472260) at lread.c:1350
#8  0x08140ef5 in Fload (file=945344908, noerror=405472260, nomessage=405472308, nosuffix=405472260, 
    must_suffix=405472308) at lread.c:895
#9  0x0813849f in Frequire (feature=408458660, filename=405472260, noerror=405472260) at fns.c:3286
#10 0x0812ebe2 in Feval (form=136773152) at eval.c:2090
#11 0x0812c7cd in Fprogn (args=0) at eval.c:424
#12 0x0812ed34 in Feval (form=1210512152) at eval.c:2037
#13 0x0812f596 in Ffuncall (nargs=2, args=0xbfffdc14) at eval.c:2731
#14 0x081570bc in Fbyte_code (bytestr=140038524, vector=1, maxdepth=-1073751024) at bytecode.c:709
#15 0x0812ebe2 in Feval (form=136775704) at eval.c:2090
#16 0x08141d84 in readevalloop (readcharfun=405588356, stream=0x858f410, sourcename=942643764, 
    evalfun=0x812e8a0 <Feval>, printflag=0, unibyte=0, readfun=405472260) at lread.c:1350
#
...

Re: random core dumps

[Richard Stallman]

The only thing I can think of to track this down
is to make compact_small_strings record details of its recent
activity in a ring buffer.  Then when it crashes you can figure
out what data led it to crash.  As you learn more, change the
history-recording code to record whatever answers you wish you knew.