Re: svn commit: r529830 - in /apr/apr/branches/1.2.x: CHANGES file_io/unix/readwrite.c test/testfile.c

[Bojan Smojver <bo...@rexursive.com>]

On Tue, 2007-04-17 at 21:58 -0500, William A. Rowe, Jr. wrote:

> I plan to at some point 'nearly close' 0.9 by creating another stable
> release.  If you wanted to backport I'd be most appreciative.

Looked at 0.9.x code and it seems that apr_file_gets() doesn't use any
locks, but relies on apr_file_read() completely (i.e. it doesn't even
have a separate buffered code path). So, I'm guessing this deadlock
isn't affecting this version.

I can still commit Joe's apr_file_gets() buffered test, if you like (I
ported that and it seems to pass). Yes?

> If you or Joe would look at
> 
> http://issues.apache.org/bugzilla/show_bug.cgi?id=41119
> 
> there is definately some evil interaction between buffering and fork()
> for exec().

OK. I'm sure Joe will have better ideas (as usual :-), but I'm sure it
won't hurt if I have a look as well.

-- 
Bojan

Re: svn commit: r529830 - in /apr/apr/branches/1.2.x: CHANGES file_io/unix/readwrite.c test/testfile.c

[Bojan Smojver <bo...@rexursive.com>]

On Wed, 2007-04-18 at 14:08 +1000, Bojan Smojver wrote:

> OK. I'm sure Joe will have better ideas (as usual :-), but I'm sure it
> won't hurt if I have a look as well.

And most certainly he did. I think it's time for folks to look at the
Bugzilla entry and maybe add a comment or two in relation to "should
this be fixed, or should it be left the way it is". As Joe pointed out,
the fix for this is easy, but whether the issue should be fixed or not
is a different matter.

http://issues.apache.org/bugzilla/show_bug.cgi?id=41119

-- 
Bojan