------- Comment #10 from ktietz at gcc dot gnu dot org 2010-04-24 12:02 -------
So I investigated this issue about mktemp in more detail and found finally the
first-scope and second-scope bug here.
Old logic was opening files and as long as mktemp failed on a second call,
things were working (the re-initialization of the template was missing, so one
of the following calls will failed). As side-effect there were many file
handles generated, which weren't closed.
Tested patch is
Index: unix.c
===================================================================
--- unix.c (revision 158676)
+++ unix.c (working copy)
@@ -889,25 +889,26 @@
template = get_mem (strlen (tempdir) + 20);
+#ifdef HAVE_MKSTEMP
sprintf (template, "%s/gfortrantmpXXXXXX", tempdir);
-#ifdef HAVE_MKSTEMP
-
fd = mkstemp (template);
#else /* HAVE_MKSTEMP */
-
- if (mktemp (template))
- do
+ fd = -1;
+ do
+ {
+ sprintf (template, "%s/gfortrantmpXXXXXX", tempdir);
+ if (!mktemp (template))
+ break;
#if defined(HAVE_CRLF) && defined(O_BINARY)
fd = open (template, O_RDWR | O_CREAT | O_EXCL | O_BINARY,
S_IREAD | S_IWRITE);
#else
fd = open (template, O_RDWR | O_CREAT | O_EXCL, S_IREAD | S_IWRITE);
#endif
- while (!(fd == -1 && errno == EEXIST) && mktemp (template));
- else
- fd = -1;
+ }
+ while (fd == -1 && errno == EEXIST);
#endif /* HAVE_MKSTEMP */
Ok for applying it to trunk, 4.5, and 4.4?
Kai
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43844
