Date: Mon, 20 Oct 2014 16:37:29 +0200
From: Markus Hoenicka <markus.hoenicka@mhoenicka.de>
Cc: Eli Zaretskii <eliz@gnu.org>, 17753@debbugs.gnu.org
I should have tried this first, but anyway: with a stack size of 1 MB
there seems to be a reproducible way to crash emacs-X11.exe, although
I
can't tell if this is just an artifact unrelated to the problem that I
reported initially. The segfault seems to be the same as previously
reported, see the attached backtrace.
This is a real stack overflow, so 1MB is evidently too low for Emacs.
(It could be that Emacs shouldn't crash in that case, but that's a
different bug, if you want to report it.)
Thread 1 (Thread 4112.0x1448):
#0 0x00000001005c2906 in ___chkstk_ms ()
at /usr/src/debug/gcc-4.8.3-3/libgcc/config/i386/cygwin.S:146
Whenever you see ___chkstk_ms at the top of the backtrace, it's a
stack overflow. That function is the one that checks for stack
overflow.
I think this "lower your stack" attempt is driving us from the real
problem, certainly with 1MB of stack.