I agree with you, Curt. But I think it's a good short term solution to help developers who are blocked with the current implementation. The solution you refer to is surely the best one, but much work will be needed to achieve it. Consequently, we could deliver it later.
Regards,
Curt Arnold wrote:
I'm uncomfortable with that solution, but I'm on deadline on something else and can't research it now. It is a problem that needs to be fixed, but I'd like to fix it by eliminating implementation details (like the config_auto.h) from the expansions of the API include files like logger.h.
On Aug 12, 2004, at 12:51 PM, Christophe de VIENNE wrote:
Cesar Eduardo Barros wrote:
On Thu, Jul 08, 2004 at 05:32:37PM -0300, Cesar Eduardo Barros wrote:
log4cxx is including a copy of the autoconf-generated config_auto.h even
when compiling another program which also uses autoconf, causing
conflicts. However, only LOG4CXX_EXPORT is really needed. I made this
quick patch to fix it.
This patch reuses the LOG4CXX define used for almost the same purpose on
config_msvc.h (to determine if we're compiling log4cxx itself or a
program which uses log4cxx).
I didn't notice some headers were using the defines from config_auto.h (HAVE_*, MUST_* and UNICODE). So, the simple solution above didn't work when I tried to use the DOM configurator.
I was trying to find a nice solution to this problem, it seems I don't have to anymore :-) Thanks !
This patch renames these defines on all headers and adds a sed-generated copy of config_auto.h prepending LOG4CXX_ to all these defines and removing the bogus PACKAGE_*, VERSION and STDC_HEADERS ones when not being included within the log4cxx sources.
This patch solves efficiently the macro problem. The idea to use sed is very good : nice job :-)
Tested compiling in the source directory, but should work with a separate object directory too.
No, but I fixed it (it was a previous issue). I also had to changed macros names in src/threadspecificdata.cpp. I'm reviewing src/*.cpp to check is there is no forgotten macro renaming.
Apart from this, I think the patch can be commited. Does everybody agree ?
Christophe
-- MichaÃl CATANZARITI log4cxx project manager
log4cxx user mailing list:
mailto:[EMAIL PROTECTED] log4cxx developer mailing list:
mailto:[EMAIL PROTECTED]
