httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@rowe-clan.net>
Subject Re: cvs commit: httpd-2.0/modules/generators mod_status.c mod_status.dsp mod_status.h
Date Thu, 24 Jul 2003 19:20:46 GMT
At 08:37 AM 7/24/2003, Joe Orton wrote:
>On Tue, Jul 22, 2003 at 08:01:43PM -0000, nd@apache.org wrote:
>> nd          2003/07/22 13:01:43
>> 
>>   Modified:    modules/generators mod_status.c mod_status.dsp mod_status.h
>>   Log:
>>   fix mod_status compilation on Win32 (stolen from mod_dav).
>>   
>>   We should consider a rename of AP_*_HOOK_* to CORE_*_HOOK_* ;-)
>
>Are these STATUS_DECLARE_* things really needed? 

Even if we use the same 'namespace', e.g. ap_, each DECLARE macro family
applies to just a single loadable module.

When we build that module (all of it's source files) we FOO_DECLARE_EXPORT 
to declare that every symbol of FOO_ is exported.  Otherwise, if you mix and
match source files, you end up with "Exported symbol was imported" sorts of
warning messages.

When we plan to link to the module (which exposes static linkage entry points),
then we provide these macros so the 3rd modules' sources know that the entry
point is imported.

Yah - there are other ways (less clean IMHO) to deal with fn entry points.
However, there is no other way to deal with data declarations.  If you mess
that up, the compiled code is not relocateable, and therefore segfaults if
another module is previously loaded at this module's expected address.

Bill



Mime
View raw message