Re: [CRASH] building xemacs, latest cvs 21.2

dv> EMACSBOOTSTRAPLOADPATH="/usr/local/src/XEmacs-21.2/src/../lisp/:/usr/local/src/xemacs" EMACSBOOTSTRAPMODULEPATH="/usr/local/src/XEmacs-21.2/src/../modules/:/usr/local/src/xemacs" ./temacs -batch -l /usr/local/src/XEmacs-21.2/src/../lisp/update-elc.el dv> Fatal error: assertion failed, file /usr/local/src/XEmacs-21.2/src/alloc.c, line 2447, abort() This ones for Olivier. First, let's have a more informative crash: Index: src/alloc.c =================================================================== RCS file: /usr/CVSroot/XEmacs/xemacs/src/alloc.c,v retrieving revision 1.42.2.51 diff -u -w -r1.42.2.51 alloc.c --- alloc.c 2000/02/21 23:07:07 1.42.2.51 +++ alloc.c 2000/02/22 12:57:29 ＠＠ -2443,8 +2443,8 ＠＠ void dumpopaque (void *varaddress, size_t size) { - if (dumpopaqueidx >= countof (dumpopaquevec)) - abort (); + assert (dumpopaqueidx < countof (dumpopaquevec)); + dumpopaquevec[dumpopaqueidx].data = varaddress; dumpopaquevec[dumpopaqueidx].size = size; dumpopaqueidx++; My guess is: Didier has a special site-load or whatever to load an unusually large number of startup elisp. Executing that elisp generates some opaques (see bytecode.c). This overflows a fixed length array. First thing I would try (after being able to reproduce the problem, which I cannot do currently), is to make all the fixed size arrays in alloc.c into dynarrs. Martin