Re: svn commit: r729579 - in /httpd/httpd/trunk: CHANGES modules/generators/mod_cgid.c

[Joe Orton <jo...@redhat.com>]

On Sat, Dec 27, 2008 at 01:12:24PM +0000, Nick Kew wrote:
> On 27 Dec 2008, at 09:52, Ruediger Pluem wrote:
>> On 12/27/2008 03:13 AM, niq@apache.org wrote:
>>> Author: niq
>>> Date: Fri Dec 26 18:13:47 2008
>>> New Revision: 729579
>>>
>>> URL: http://svn.apache.org/viewvc?rev=729579&view=rev
>>> Log:
>>> PR#39332: fix for segfault problem with mod_cgid on Solaris
>>> Patch by Masaoki Kobayashi
>>>
>>> Modified:
>>>     httpd/httpd/trunk/CHANGES
>>>     httpd/httpd/trunk/modules/generators/mod_cgid.c
>>
>> Just for my understanding. This patch works around a bug in Solaris 
>> with write and UNIX domain sockets, right? Otherwise I wouldn't get 
>> why the new code using writev would be better thatn the old one using 
>> write.
>
> It's a bug with a number of reports, though I haven't actually been able
> to reproduce it myself on opensolaris.
>
> Rationale for the patch is that I reviewed it and ruled out the risk of a
> regression(!)  

Ah, a zero-risk-of-regression patch, I'll just quote you on that :)

Given that as Justin mentioned, there is already a configure check to 
avoid building cgid on Solaris systems which have the bug in question, I 
think this leaves unresolved questions:

1) is the configure check not sufficient to catch the set of systems 
which have a kernel bug?

2) is there some other bug (cgid or Solaris or whatever) at large here?

I'd rather not see this applied unless this is sorted out, since it 
clearly does *not* have a zero risk of regression.

Regards, Joe