Ryan Bloom <[EMAIL PROTECTED]> writes:

> I am missing something here.  This patch requires us to use the non-portable
> W* macros in Apache.  That is wrong.  

This patch does not force an APR app to use any W* macros other than
WCOREDUMP(), and most apps don't care about that.

>                                This patch does not help APR to be
> more portable, rather it allows programs that use APR to continue to use
> non-portable constructs to solve the problem.  I would fully expect that in
> order for this to be done correctly, we will need to modify 
> ap_process_child_status in mpm_common.c to not use the W* functions.  If we
> do anything else, then we have not helped APR to be portable.

We're trying to make an APR app more portable.  I certainly plead
guilty to allowing an APR app to use a non-portable construct when the
app considers it necessary.  I hardly invented the concept.  It is an
important feature of APR in various areas.

What this patch does, if I may offer my version of
motherhood-and-apple-pie, is increase the ability of an APR app to do
meaningful stuff portably when compared with current CVS.  It allows
an APR app to do non-portable stuff if it so chooses by making
available the native status code.

As far as apr_wait_or_timeout() and apr_process_child_status(), please
see my my previous patch for examples on how to reduce the amount of
non-portable constructs in Apache.  I just don't want to muck with
that at the moment.  Unlike in my previous patch, I did not choose to
use the portable view of exit status whenever possible because I want
to get APR fixed appropriately first (while allowing all MPMs to work
correctly).

-- 
Jeff Trawick | [EMAIL PROTECTED] | PGP public key at web site:
       http://www.geocities.com/SiliconValley/Park/9289/
             Born in Roswell... married an alien...

Reply via email to