On Fri, May 26, 2006 at 09:33:12PM +0000, Eric Blake wrote:
>cgf wrote:
>
>> >0022FF88 61005EB3 (00000000, 00000000, 00000000, 00000000)
>>
>> That's apparently a call to mktemp, although I can't tell what's causing
>> the problem.
>
>Well, looking at the source to the cygwin distribution of texinfo, it is
>pretty obvious:
>
>
>$ diff -u texindex.c.orig texindex.c
>--- texindex.c.orig 2004-04-11 11:56:47.001000000 -0600
>+++ texindex.c 2005-10-07 08:43:50.001000000 -0600
>@@ -99,6 +99,9 @@
> /* Directory to use for temporary files. On Unix, it ends with a slash. */
> char *tempdir;
>
>+/* Basename for temp files inside of tempdir. */
>+char *tempbase;
>+
> /* Number of last temporary file. */
> int tempcount;
>
>@@ -190,6 +193,11 @@
>
> decode_command (argc, argv);
>
>+ /* XXX mkstemp not appropriate, as we need to have somewhat predictable
>+ * names. But race condition was fixed, see maketempname.
>+ */
>+ tempbase = mktemp ("txidxXXXXXX");
>+
>...
>
>OOPS!!!!
Calm down, please.
>The texinfo maintainer, in their modifications to upstream texindex, is
>passing a READ-ONLY string to mktemp, and is deserving of their crash
>(although cygwin could perhaps handle it more gracefully, perhaps by
>returning NULL on a fault so that at least the stack trace is not in
>the middle of cygwin1.dll). But if you are going to go to the effort
>of modifications, use mkstemp instead (mktemp is insecure, after all).
>I would further argue that maketempname should use the O_BINARY flag to
>open(), unless texindex can handle text mounts gracefully. Can we get
>a new texinfo release soon with this fixed?
This patch is actually not mine. I don't remember exactly where it came
from, but google shows a similar patch circulating in a couple of
distributions.
cgf
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
