perl-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Steve Hay" <>
Subject RE: [httpd24 branch] merge with trunk?
Date Mon, 15 Jul 2013 12:58:34 GMT
Jan Kaluža wrote on 2013-07-15:
> On 07/15/2013 12:52 PM, Steve Hay wrote:> Jan Kaluža wrote on 2013-07-
> 15:
>>> On 07/15/2013 12:30 PM, Steve Hay wrote:
>>>> Jan Kaluža wrote on 2013-07-15:
>>>>> On 07/15/2013 10:48 AM, Steve Hay wrote:
>>>>>> So I now have a build of mod_perl against httpd-2.4. Yay!
>>>>>> However, it doesn't run yet... it's crashing when starting up
> the  >>>> server. I will look into that very soon...
>>>>> It crashed for me too here. It used to work before the changes
> you've  >>>> done in last week. I will review them and try to find out
> what's wrong.
>>>> That's good, actually, because it simultaneously means it isn't a
>>>> Windows-only failure and greatly narrows down where the problem
> must  >>> be :-)  >>>  >>> I will look soon too if you haven't
> me to it ;-)  >>>  >>  >> I think I have fixed that (together with
> some compiler warnings) in  >> r1503171. Hopefully it will still
> compile properly on Windows. Please  >> inform me if this revision
> works for you. All expected tests are  >> passing on Fedora in this
> revision again.
>> At a quick glance I think it won't compile again because dTHXa and
> dSP are declarations, which are now mixed up with code again.
> You are right. Maybe something like r1503193 could workaround that?
> Although whole situation around that code is strange (see my FIXME
> comment in the mentioned revision).

It now builds on Windows again.

I've just rearranged it very slightly again in r1303215 to put back the early return if the
key is not found, like you had done in r1503171. I've also protected the modperl_interp_pool_select()
call and dTHXa() declaration with #ifdef USE_ITHREADS like other callers do.

Not sure what to say about the FIXME comment, but if modperl_interp_pool_select() can now
return NULL then the other callers should test for that too, rather than always blindly dereferencing
the return value.

I will look at that shortly, but right now I still have it crashing on startup (with or without
revision r1303215), so we obviously didn't have the same crashing problem after all :-/ I
will try to fix that next.

To unsubscribe, e-mail:
For additional commands, e-mail:
View raw message