Re: [PATCH] HAVE_MKSTEMP tries to kill not existent tmp-file in $HOME and $TMPDIR (1.9.2)

On Thu, May 24, 2007 at 06:25:06PM +0200, Rado S wrote:
> Moin,
> 
> after revisiting this issue, I offer this attached patch:
> even though this 5s delay for each successful posting is very
> annoying, the debug output is totally useless, when you read "man
> mkstemp", because it says:
> ------ QUOTE BEGIN ------
> The function thus prevents any possible race condition between
> testing whether the file exists and opening it for use.
> ------- QUOTE END -------

sounds like a solaris thing as neither the linux nor the freebsd or the
opengroup manpages mention this (currently I have no solaris box)

> Therefore by defintion there any ENOENT case is just a side-effect:
> the idea of this function is to not have the tmp-file exist.

returning ENOENT in case of "No such file" is stupid in this case and if the
dir doesnt exist it should return ENOTDIR. the current code is:

	fd = mkstemp(filename);
#   ifdef DEBUG
	sverrno = errno;
	if (fd == -1 && sverrno)
		wait_message(5, "HAVE_MKSTEMP %s: %s", filename, strerror(sverrno))
#   endif /* DEBUG */

which should suffice. no special case for ENOENT (as this sounds like a
broken implementation) and no user notify when n ot running in debug mode
(as this is not a grave error.