[xemacs-hg @ 2002-05-23 11:46:08 by ben] fix for raymond toy's crash, alloca crashes, some recover-session improvements files.el: Recover-session improvements: Only show session files where some files can actually be recovered, and show in chronological order. subr.el, menubar-items.el: As promised to rms, the functionality in truncate-string-with-continuation-dots has been merged into truncate-string-to-width. Change callers in menubar-items.el. select.el: Document some of these funs better. Fix problem where we were doing own-clipboard twice. Makefile.in.in: Add alloca.o. Ensure that alloca.s doesn't compile into alloca.o, but allocax.o (not that it's currently used or anything.) EmacsFrame.c, abbrev.c, alloc.c, alloca.c, callint.c, callproc.c, config.h.in, device-msw.c, device-x.c, dired.c, doc.c, editfns.c, emacs.c, emodules.c, eval.c, event-Xt.c, event-msw.c, event-stream.c, file-coding.c, fileio.c, filelock.c, fns.c, glyphs-gtk.c, glyphs-msw.c, glyphs-x.c, gui-x.c, input-method-xlib.c, intl-win32.c, lisp.h, lread.c, menubar-gtk.c, menubar-msw.c, menubar.c, mule-wnnfns.c, nt.c, objects-msw.c, process-nt.c, realpath.c, redisplay-gtk.c, redisplay-output.c, redisplay-x.c, redisplay.c, search.c, select-msw.c, sysdep.c, syswindows.h, text.c, text.h, ui-byhand.c: Fix Raymond Toy's crash. Repeat to self: 2^21 - 1 is NOT the same as (2 << 21) - 1. Fix crashes due to excessive alloca(). replace alloca() with ALLOCA(), which calls the C alloca() [which uses xmalloc()] when the size is too big. Insert in various places calls to try to flush the C alloca() stored info if there is any. Add MALLOC_OR_ALLOCA(), for places that expect to be alloca()ing large blocks. This xmalloc()s when too large and records an unwind-protect to free -- relying on the caller to unbind_to() elsewhere in the function. Use it in concat(). Use MALLOC instead of ALLOCA in select-msw.c. xemacs.mak: Add alloca.o.

/* Synched up with: FSF 19.31. */

#include "sunos4-0.h"

/* 4.1.1 makes these system calls interruptible.  */

#define INTERRUPTIBLE_OPEN
#define INTERRUPTIBLE_CLOSE
#define INTERRUPTIBLE_IO

/* Cause the compilation of oldxmenu to use the right -I option.  */
#define OLDXMENU_OPTIONS CFLAGS=C_SWITCH_SYSTEM

#if 0 /* RMSmacs */
FSF says the following, but apparently the HAVE_RES_INIT stuff takes
care of it properly?
/* #if 0 This isn't right.  Apparently some sites do have -lresolv
	 but don't use that.  On those systems, the code below loses.
	 There's no way to win automatically unless someone
	 figures out a way of determining automatically which way is right
	 on any given system.  */
#endif

/* Some systems do not run the Network Information Service, but have
   modified the shared C library to include resolver support without
   also changing the C archive library (/usr/lib/libc.a).  If we
   detect the presence of libresolv.a, use -lresolv to supplement libc.a.

   We used to have #ifdef HAVE_GETHOSTNAME is to prevent configure from
   setting libsrc_libs to -lresolv in lib-src/Makefile.  But nowadays
   configure is smarter about computing libsrc_libs, and would not
   be fooled.  Anyway, why not use -lresolv in lib-src?  */
/* #ifdef HAVE_GETHOSTNAME */
/* XEmacs change: -lresolve should be added only if we have RES_INIT,
   not if we don't */
#ifdef HAVE_RES_INIT
#define LIBS_SYSTEM "-lresolv"
#endif
/* #endif */

#if 0 /* Not necessary, since SYSTEM_MALLOC is defined in sunos4-0.h.  */
/* Tell GNU malloc to compensate for a bug in localtime.  */
#define SUNOS_LOCALTIME_BUG
#endif

/* Define dlopen, dlclose, dlsym.  */
#define USE_DL_STUBS

/* This appears to be broken on SunOS4.1.[123] */
#define BROKEN_SIGIO