It too suffers from the same problem as operator==: Compares name
pointers, not name strings. Thus with two things to patch, had to
refactor it.
(before() doesn't seem to get called when running the bridgetest at
least, though.)
Noticed that there is no need to wrap the replacement asm snippets
with dummy C functions. Just having them at the outermost level in the
source file works as expected.
A compile time check ensures the common case of streaming just a plain
C-style string literal still produces reasonably compact call-site code.
The format-string variants are still available in sal/detail/log.h, but
only to be used in obsolete osl/diagnose.h etc., and going to be removed
again eventually.
So that variables only used within SAL_INFO/WARN do not lead to
unused variable warnings in product builds. Also for OSL_TRACE
(which used to work that way before already); allowing to revert
a recent warning fix.
* New SAL_INFO..., SAL_WARN... macros.
* New SAL_STREAM supersedes OSL_FORMAT.
* oustringostreaminserter.hxx moved from unotest to rtl (and always UTF-8 now).
* TODO to enable GCC __attribute__((format)) in sal/log.h (requires call-site
cleanup).
* Further functionality in tools/debug.hxx (DBG_MEMTEST, DBG_CTOR, etc.) not yet
addressed.
* Some replacements tools String -> rtl::OUString.
You are not expected to be make'ing here without knowing what you are
doing anyway. Just unconditionally require the stuff needed by the
sample cppunit tests mentioned on the last lines.
Now it runs until a crash caused by ICU library mixup between our ones
and the system ones, they have the same names but aren't
compatible... (see 7ee03666d7cfc27453315e1682e3c32d39a031b6).
Add possibility to start strace tracing the process. Unfortunately it
this doesn't seem to work as nicely as one might have hoped. If the
process crashes, the last strace output does not show up. Some
buffering that gets abruptly discarded when the straced process dies?
Add a function to the native code to look up an archive member in the
.apk and return a pointer to it. To be used for non-compressed
resources only. We mmap the whole .apk. The Zip format parsing code is
borrrowed from Mozilla's APKOpen.cpp.
Correspondingly, add to the local build.xml a re-definition of the
"-package-resources" target from Ant's build.xml modified to not
compress resources.
Improved the Makefile a lot. New target "install" to build the apk and
install it without constructing the apk twice.
Other minor changes here and there.
Android's dladdr() stores just the basename of a shared object in
Dl_info::dli_fname. LibreOffice expects either a complete absolute
pathname or a relative path from the current directory to be returned
there. So look up the full pathname of the library from
/proc/self/maps.
Also remove the fallback message loop after lo_main() returns, instead
just exit.
