open_temp_exec_file_mnt bug with getmntent_r

[Andrew Haley]

On 09/03/2010 10:34 AM, Mark Wielaard wrote:
> On Fri, 2010-08-27 at 11:37 +0200, Mark Wielaard wrote:
>> On Fri, 2010-08-20 at 13:05 +0200, Mark Wielaard wrote:
>>> Probably nobody noticed before because earlier open_temp_exec_file_opts
>>> (TMPDIR, /tmp, /var/tmp, /dev/shm, HOME) always allow creating files.
>>> But if you have a weird setup that doesn't, then the
>>> open_temp_exec_file_mnt mount point search won't help you because the
>>> getmntent_r () return check is wrong. It fails when an entry is found
>>> instead of when no entry is found (NULL is returned).
>>>
>>> This fixes that case:
>>>
>>> 2010-08-20  Mark Wielaard  <mjw@redhat.com>
>>>
>>>         * src/closures.c (open_temp_exec_file_mnt): Check if getmntent_r
>>>         returns NULL.
>>>
>>> You can also test this by commenting out the earlier 
>>> open_temp_exec_file_opts options.
>>
>> Ping. Any comments?
> 
> Adding java-patches to see if anybody there thinks this is a good idea.
> Attached patch is against gcc/libffi.

It's obviously wrong.  Please commit this.

Andrew.