Re: [Bug: 21.5-b10] --use-kkcc + gnus = core

ms> That looks like the usual current problem with kkcc, namely ms> that it uses more stack than the normal collector. That's right. Basically, s/--use-kkcc + //. CFLAGS += -DREGEX_MALLOC probably wins for you. (-: Gnus is evil and always has been. I wouldn't have it any other way. It's the best regression test we've got. :-) Here the problem is very likely the regexp allocation scheme. Regexps can be highly memory-intensive if you use backtracking (expressions involving * are typical) and unlimited search in a big buffer. Gnus does this a lot (note that the big buffer might be something as simple as a short spam or virus mail containing lots of images or an executable). On the other hand, typically they're not, and thus we don't hang on to the backtrack caches, we reallocate them on every invocation of the internal regexp matcher. So for efficiency alloca is used. It seems that in the distant past many allocas would return 0 on a failure, but modern allocas (in particular gcc's and glibc's) generally make no attempt to let you know about a failure. If I understood his oblique comment correctly, Ben claims to have tried to do something about this in 21.5, but not successfully, as I crashed on a long error message from my 21.4-oriented init.el. -DREGEX_MALLOC fixed the regexp crash. -- Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Ask not how you can "do" free software business; ask what your business can "do for" free software.