Last updated: 12 Nov 2006
- new log4j naming conventions: category -> logger
- multiple appenders per category
- improve errors in configuration file parser
- location_info is broken (need macro ?)
- remove initialization of log4c when library is loaded (compatibility
. Currently printing some reminder messages in the init and fini sections with gcc.
I guess we ditch those just before the release of 1.2 ? Or always leave em in the
- add void* to location_info
- thread safe ?
. see the "Log4C Developers Guide" for a discussion of the issues
and some sugggestions for fixing them.
. behavior with autoconf is to use expat if it's there. That's ok but
there needs to be an option so that prevents it being picked up--otherwise
it's hard for release engineers to control what they manufacture with Log4C.
- what is the stauts of the macosx build files--do they work with the latest workspace ?
- improve sd/error.c
. currently the sd_debug are functions...is that costly in the optimized build even if
the body is empty ?
- docs needs an update
. home web page needs to be synced up to 1.1.0, link to latest API doc,
and the draft developers guide.
. Could probably use a "Log4C Internals" doc as well
- rolling file appender:
. take a file lock as well as the mutex. That would protect
the file from other processes that might access it.
. the mapping of pthread to the native Windows thread functrions is done
by macros definined in sd_xplatform.h. This works ok...but the join/cancel macros
are not returning the right values.
. two of the tests in the test_rc program are failing...need to look at that.
. In category.h we make a special case for some of the inline functions on windows.
This is not really required as the __inline (or __forceinline) directoves exist on windows.
So we could just define inline to be __inline in sd_xplatform.h.
. write a test program focused on lifecycle management of log4C objects.
For example it can open, append, close in a loop to make sure everything works and
ther eare no memory leaks. What about log4c_init, log4c_fini in a loop...not sure what that
gives...not ethat the base types are not freed by log4c_fini....