updated.

[platform/upstream/glib.git] / ChangeLog

diff --git a/ChangeLog b/ChangeLog
index c38b90e..c967539 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,48 @@
+2000-04-07  Sebastian Wilhelmi  <wilhelmi@ira.uka.de>
+
+       * grand.c (g_rand_new): Fixed bug. Thanks to Marko Kreen
+       <marko@l-t.ee> for reporting that.
+
+2000-03-26  Tor Lillqvist  <tml@iki.fi>
+
+       * README.win32: Tell about using the mingw-based gcc, which is
+       much easier than modifying the cygwin gcc to product mingw code
+       for the msvcrt runtime.
+
+       * makefile.cygwin.in (WIN32APIHEADERS): Kludge to make it work
+       with a "pure" mingw gcc, too.
+
+2000-03-24  Sebastian Wilhelmi  <wilhelmi@ira.uka.de>
+
+       * garray.c: Made GArray behave correctly. Now zero_terminated
+       really means, that the element array->data[array->len] exists and
+       is zeroed, and clear means that any unassigned elements obtained
+       through g_array_set_size (the only way to get unassigned elements
+       AFAICT) are zeroed. Added some macros to make the code more
+       obvoius. Also made GPtrArray zero elements after
+       g_ptr_array_set_size. This is done in a portbale way (assignment
+       of NULL instead of just memsetting it to zero), though that might
+       be more portability than we actually want.
+
+       * Makefile.am, gthread/Makefile.am, gmodule/Makefile.am,
+       tests/Makefile.am: Added various win32 related *.in files to
+       EXTRA_DIST to let 'make distcheck' procude all the corresponding
+       files, which it silently fails to do currently.
+
+2000-03-23  Sebastian Wilhelmi  <wilhelmi@ira.uka.de>
+
+       * configure.in: After finding the right thread library (containing
+       e.g. pthread_create) we now search for the right realtime library
+       (containing e.g. sched_get_priority_max). Makes the output of the
+       thread related libraries correct. 
+
+       * gtimer.c (g_usleep): The current implementation of g_usleep
+       (simply calling select) doesn't work reliable for multi-threaded
+       programs on some platforms (bad omen for the main loop....), so I
+       changed the implementation for thread-using programs to wait for a
+       GCond for the specified amount of time (NB: sleep and usleep are
+       not MT-safe in general, because they often use signals).
+
 Wed Mar 22 16:49:57 2000  Owen Taylor  <otaylor@redhat.com>
 
        * gmem.c (g_mem_chunk_area_compare): Fix indentation.