Re: [AC21.5R21.4] darwin: configure pdump fix and DLSYM_NEEDS_UNDERSCORE

redirecting thread to xemacs-beta Andrew> However, something in regexp.c is sure broken. Here's the Andrew> crash I get when starting xemacs when missing a needed Andrew> file in a load statement in my init.el file. Running Andrew> xemacs -q eliminates the segfault. Looks like the known "not so big Mac stack" problem. See PROBLEMS for details, but the gist is that the standard 512kB stack is way too small, and setting it to 8MB is suggested. If you have already adjusted your stack, and this is something different, please say so! A patch to increase the stack size XEmacs requests automatically at run time has been suggested, but this is just plain wrong. The problem is the use of alloca for allocating the "failure stack" in the regexp interpreter. The real problem is not that alloca limits size, it's that alloca has only one failure mode on many platforms: stack corruption. Eventually we'll see a failure, and alloca guarantees that it will only be indirectly diagnosable, and not preventable. This needs to be fixed, by switching from alloca to malloc at some fixed level. Searches that use more than 1kB (one KILObyte) are rare ---it's possible that switching to malloc when the stack size hits 1k would have negligible performance impact. Demanding an extra 7.5MB for the rare case seems absurd, especially since it's not always enough. -- 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.