Logged as LOGCXX-169 (https://issues.apache.org/jira/browse/
LOGCXX-169) and resolved. Basically isolated XMLLayoutTestCase from
the implementation of __LOG4CXX_FUNC__.
I added LOGCXX-170 as a distinct bug which would move the sniffing of
method name macros to the configure step and add sniffing of __func__
to the list of potential sources for location info. If anyone is
interested in taking a shot, please feel free to attach a patch.
I've set up a Solaris build environment and have been doing a little
more digging on wchar_t on the platform. wchar_t on Solaris is
locale dependent. It is UTF-32 when the locale uses UTF-8, but can
vary on other locales. On Mac's, it is never explicitly defined.
Currently the lack of a defined transform from wchar_t to UTF-8 is
reflected by a link failure when attempting to compile log4cxx with
wchar_t support on those platforms. While it is less than ideal, the
only way I see forward is to use wcstombs to convert from wchar_t to
the current multibyte set and then use apr_xlate to convert from the
current multibyte set to UTF-8. If the current multibyte set is
UTF-8, then last step is redundant. If it is not, then some wchar_t
values might not be convertable to the current multibyte set and
would be replaced with a "?" even though they could be represented in
UTF-8 if a single step translation was available.
