MB> What is the effect? Here's one bug: MB> with pdump: MB> (martin＠polgar) /usr/local/x/build/mpps/src $ ./xemacs -batch -eval '(message "%s %s" (make-temp-name "foo") (make-temp-name "foo"))' MB> fooWdDWXY fooWdDWXY MB> without pdump: MB> (martin＠polgar) /usr/local/x/build/mps/src $ ./xemacs -batch -eval '(message "%s %s" (make-temp-name "foo") (make-temp-name "foo"))' MB> fooXdD_XY fooXdDgqm The bug with Frunning_temacs_p() still needs to be fixed (I cast a vote for Olivier to do it), but here's a fix for this specific problem, with other "improvements". src/ChangeLog: 2000-03-10 Martin Buchholz <martin(a)xemacs.org&gt; * emacs.c: Add reinit_vars_of_fileio. * symsinit.h: Add reinit_vars_of_fileio. * fileio.c (reinit_vars_of_fileio): New. * fileio.c (Fmake_temp_name): Initialize temp_name random number from microseconds to make collisions even less likely. Initialize always at process startup time. (make-temp-name) used to return the same file name twice in a row when PDUMP. Random stylistic fiddling. Comment fixes. tests/ChangeLog: 2000-03-10 Martin Buchholz <martin(a)xemacs.org&gt; * automated/mule-tests.el: Check that (make-temp-name) returns unique values. Index: src/symsinit.h =================================================================== RCS file: /usr/CVSroot/XEmacs/xemacs/src/symsinit.h,v retrieving revision 1.31.2.22 diff -u -w -r1.31.2.22 symsinit.h --- symsinit.h 2000/02/05 07:09:02 1.31.2.22 +++ symsinit.h 2000/03/11 00:59:58 ＠＠ -290,6 +290,7 ＠＠ void reinit_vars_of_extents (void); void vars_of_faces (void); void vars_of_fileio (void); +void reinit_vars_of_fileio (void); void vars_of_filelock (void); void vars_of_floatfns (void); void vars_of_font_lock (void); Index: src/fileio.c =================================================================== RCS file: /usr/CVSroot/XEmacs/xemacs/src/fileio.c,v retrieving revision 1.51.2.22 diff -u -w -r1.51.2.22 fileio.c --- fileio.c 2000/03/05 21:55:47 1.51.2.22 +++ fileio.c 2000/03/11 01:00:09 ＠＠ -635,8 +635,10 ＠＠ This implementation is better than what one usually finds in libc. --hniksic */ +static unsigned int temp_name_rand; + DEFUN ("make-temp-name", Fmake_temp_name, 1, 1, 0, /* -Generate temporary file name starting with PREFIX. +Generate a temporary file name starting with PREFIX. The Emacs process number forms part of the result, so there is no danger of generating a name being used by another process. ＠＠ -645,8 +647,9 ＠＠ be an absolute file name. */ (prefix)) +{ + static const char tbl[64] = { - static char tbl[64] = { 'A','B','C','D','E','F','G','H', 'I','J','K','L','M','N','O','P', 'Q','R','S','T','U','V','W','X', ＠＠ -654,13 +657,12 ＠＠ 'g','h','i','j','k','l','m','n', 'o','p','q','r','s','t','u','v', 'w','x','y','z','0','1','2','3', - '4','5','6','7','8','9','-','_' }; - static unsigned count, count_initialized_p; + '4','5','6','7','8','9','-','_' + }; Lisp_Object val; Bytecount len; Bufbyte *p, *data; - unsigned pid; CHECK_STRING (prefix); ＠＠ -686,40 +688,33 ＠＠ /* VAL is created by adding 6 characters to PREFIX. The first three are the PID of this process, in base 64, and the second three are - incremented if the file already exists. This ensures 262144 - unique file names per PID per PREFIX. */ + a pseudo-random number seeded from process startup time. This + ensures 262144 unique file names per PID per PREFIX per machine. */ - pid = (unsigned)getpid (); - *p++ = tbl[pid & 63], pid >>= 6; - *p++ = tbl[pid & 63], pid >>= 6; - *p++ = tbl[pid & 63], pid >>= 6; + { + unsigned int pid = (unsigned int) getpid (); + *p++ = tbl[(pid >> 0) & 63]; + *p++ = tbl[(pid >> 6) & 63]; + *p++ = tbl[(pid >> 12) & 63]; + } /* Here we try to minimize useless stat'ing when this function is invoked many times successively with the same PREFIX. We achieve - this by initializing count to a random value, and incrementing it - afterwards. */ - if (!count_initialized_p) - { - count = (unsigned)time (NULL); - /* Dumping temacs with a non-zero count_initialized_p wouldn't - make much sense. */ - if (NILP (Frunning_temacs_p ())) - count_initialized_p = 1; - } + this by using a very pseudo-random number generator to generate + file names unique to this process, with a very long cycle. */ while (1) { struct stat ignored; - unsigned num = count; - p[0] = tbl[num & 63], num >>= 6; - p[1] = tbl[num & 63], num >>= 6; - p[2] = tbl[num & 63], num >>= 6; + p[0] = tbl[(temp_name_rand >> 0) & 63]; + p[1] = tbl[(temp_name_rand >> 6) & 63]; + p[2] = tbl[(temp_name_rand >> 12) & 63]; /* Poor man's congruential RN generator. Replace with ++count for debugging. */ - count += 25229; - count %= 225307; + temp_name_rand += 25229; + temp_name_rand %= 225307; QUIT; ＠＠ -1846,7 +1841,7 ＠＠ { return Fsignal (Qfile_error, list3 (build_translated_string ("Creating directory"), - build_translated_string ("pathame too long"), + build_translated_string ("pathname too long"), dirname_)); } strncpy (dir, (char *) XSTRING_DATA (dirname_), ＠＠ -4321,4 +4316,26 ＠＠ #else Vdirectory_sep_char = make_char ('/'); #endif + + reinit_vars_of_fileio (); +} + +void +reinit_vars_of_fileio (void) +{ + /* We want temp_name_rand to be initialized to a value likely to be + unique to the process, not to the executable. The danger is that + two different XEmacs processes using the same binary on different + machines creating temp files in the same directory will be + unlucky enough to have the same pid. If we randomize using + process startup time, then in practice they will be unlikely to + collide. We use the microseconds field so that scripts that start + simultaneous XEmacs processes on multiple machines will have less + chance of collision. */ + { + EMACS_TIME thyme; + + EMACS_GET_TIME (thyme); + temp_name_rand = (unsigned int) (EMACS_SECS (thyme) ^ EMACS_USECS (thyme)); + } } Index: src/emacs.c =================================================================== RCS file: /usr/CVSroot/XEmacs/xemacs/src/emacs.c,v retrieving revision 1.82.2.53 diff -u -w -r1.82.2.53 emacs.c --- emacs.c 2000/03/06 20:57:45 1.82.2.53 +++ emacs.c 2000/03/11 01:00:16 ＠＠ -1264,7 +1264,7 ＠＠ using a global variable that has been initialized earlier on in the same function - Any of the object-creating functions on alloc.c: e.g. + Any of the object-creating functions in alloc.c: e.g. make_pure_*() make_string() ＠＠ -1680,6 +1680,7 ＠＠ reinit_vars_of_event_stream (); reinit_vars_of_events (); reinit_vars_of_extents (); + reinit_vars_of_fileio (); reinit_vars_of_font_lock (); reinit_vars_of_glyphs (); reinit_vars_of_glyphs_widget (); Index: tests/automated/mule-tests.el =================================================================== RCS file: /usr/CVSroot/XEmacs/xemacs/tests/automated/Attic/mule-tests.el,v retrieving revision 1.1.2.6 diff -u -w -r1.1.2.6 mule-tests.el --- mule-tests.el 2000/02/19 09:25:03 1.1.2.6 +++ mule-tests.el 2000/03/11 01:00:17 ＠＠ -299,6 +299,7 ＠＠ (file-name-coding-system 'iso-8859-2)) ;; This is how you suppress output from `message', called by `write-region' (flet ((append-message (&rest args) ())) + (Assert (not (equal name1 name2))) (Assert (not (file-exists-p name1))) (write-region (point-min) (point-max) name1) (Assert (file-exists-p name1))

On Wed, 15 Mar 2000 07:36:28 EST, Olivier Galibert said: I believe somebody said that the portable dumper will allow using purify on a dumped emacs, so it may in the future be possible to be running purify in an xemacs? -- Valdis Kletnieks Operating Systems Analyst Virginia Tech

On Wed, 15 Mar 2000 10:06:19 EST, Olivier Galibert said: Make note: Ingest all required caffiene, THEN reply to mail. Don't reply to mail after first 8 ounces if 16 are needed for proper operation. ;) /Valdis