On Friday, 20 December 2002, Stephen J. Turnbull wrote:
>>>>> "sjt" == Stephen J Turnbull
<stephen(a)xemacs.org> writes:
sjt> --use-kkcc --mule XEmacs dumps core in make check,
sjt> mule-tests.el for me. With over 37000 stack frames, I think
sjt> the SEGV is simply stack exhaustion. I think mule-tests.el
sjt> may create a large list of characters from a string and/or a
sjt> char-table, a unicode char-table could easily have 10s of
sjt> thousands of elements.
I checked the backtrace and it's the first test (char-test) in
mule-tests, which creates a list of all characters known to Mule,
which I would guess is in the neighborhood of 60,000, and could be as
many as 100,000. That's a long list.
Indeed, if I increase the stack enough, the test succeeds. The
default stack is apparently 2044kB; 16000kB wasn't enough, 32000kB
was. wo-wo-woah!
Probably unrelated. I got a stack overflow error (no crash) when I ran
compile in a directory with a lot of output from the compiler. I guess the
font-lock regexp parsing of the compiler output crashes when the
compilation buffer becomes huge. Do you know of anything that I can do to
prevent this ? This was with XEmacs-21.4.10 on Linux/x86.
-kitty.