2001-01-25 Tor Lillqvist <tml@iki.fi>
* giowin32.c: Socket support rewritten. It was utterly broken, and
untested in fact. We still do use a thread for each socket being
watched, but instead of blocking in recv() (which of course was
plain stupid for sockets being liste()ed on), we block in
select(). The read method for sockets calls recv(). It is now
possible for the application to call accept(), recv() or send() in
the callback, just like on Unix. Tested with code kindly provided
by Andrew Lanoix.
Rename g_io_channel_win32_new_stream_socket() to
g_io_channel_win32_new_socket() as it isn't restricted to stream
sockets.
* gmain.c (g_poll): Related changes in the Win32 version of
g_poll(). When polling for messages, always do a PeekMessage()
first. We used to miss messages if several were posted between
calls to g_poll().
* giochannel.h: Improve Win32-related comments.
* gutf8.c: (Win32) Include <stdio.h> for sprintf.
* tests/gio-test.c: (Win32) Add tests for polling for Windows
messages.
* tests/makefile.mingw.in: Remove superfluous compilation command
line.
Wed Jan 3 14:10:49 2001 Owen Taylor <otaylor@redhat.com>
* gmain.[ch]: Switch GMainLoop to be ref/unref, use to
make dropping reference to running loop safe.
Wed Dec 13 20:41:49 2000 Owen Taylor <otaylor@redhat.com>
* gmain.c (g_source_unref_internal): Unref callback->cb_data
if it was still set when the source is freed. (Usually, this
will be done by g_source_destroy.)
2000-12-14 Tor Lillqvist <tml@iki.fi>
* giowin32.c (g_io_win32_dispatch): Warn if no callback. Call
callback correctly.
(g_io_win32_create_watch): Fix typo.
(g_io_win32_fd_create_watch): Ditto.
(g_io_channel_unix_new): If it is a file descriptor (i.e., a Unix
fd lookalike provided by the C library), call
g_io_channel_win32_new_fd(). If it is a socket (from WinSock),
call g_io_cahnnel_win32_new_stream_socket(). Hopefully sockets and
fds don't overlap. TODO: Implement also datagram sockets.
(g_io_channel_win32_poll): Call g_main_context_get_poll_func().
* gcompletion.h: Include <unistd.h> only on Unix. Is this
inclusion really needed here? OTOH, do include <stddef.h>, for
size_t.
* gmessages.c: (Win32) Don't define a function called "write" that
might clash with the prototype from <io.h>, use a #define.
* glib.def: Update.
* gmain.c (g_source_add_poll): Don't return a value from void
function.
(g_main_context_get_poll_func): Compile also for non-Win32, as
presumably was intended. The result var is a GPollFunc, not a
GPollFunc*. Return the result!
gobject:
2000-12-14 Tor Lillqvist <tml@iki.fi>
* makefile.mingw.in: Update, include parts from Makefile.am to
build gmarshal.[ch]. Some day, we won't need these separate
makefiles for Win32 compilation. I hope.
* makefile.msc.in: Update. No use trying to build gmarshal.[ch]
here, it would require Unixish tools. MSVC users building from CVS
sources are out of luck.
* gobject.def: Update.
* gmain.c, gmain.h (g_main_context_new, g_main_context_destroy): GMainContext useful
in implementing some additional styles of main loop usage. To do this, however, Joe
Hacker needs to be able to create/destroy GMainContext's at will. This is just an
export of existing functionality, rather than new functionality.
They are listed in the "Low level functions for implementing custom main loops"
section of the header file, to avoid confusing people.
* gobject/Makefile.am:
. You have to 'touch oldest-source-stamp' if you want to avoid having
the Makefile constantly rebuild itself.
. Fix marshaller generation rules to work with srcdir != builddir
(there were issues with trying to run "./glib-genmarshal", etc.)
Sun Dec 10 10:47:11 2000 Owen Taylor <otaylor@redhat.com>
* gmain.c (g_source_destroy_internal): Remove pollfds
from the context here, not when actually freeing the
source.
* gmain.c (g_source_unref_internal): Free source list
and source, call source->source_funcs->destroy().
* giochannel.c: Unreference io_channel properly.
Thu Dec 7 15:22:30 2000 Owen Taylor <otaylor@redhat.com>
* tests/mainloop-test.c (recurser_start): Add a bunch
of unrefs.
* gmain.c (g_source_attach): Reference the source
when adding (pointed out by Elliot)
Tue Dec 5 12:23:04 2000 Owen Taylor <otaylor@redhat.com>
* gmain.[hc]: Major change in API for creating sources
to handle multiple main loops (GMainContext *).
GSources are now exposed as GSource * and implemented
with structure derivation.
* giochannel.[ch]: Changed vtable for GIOChannel to correspond
to the new mainloop API, add g_io_channel_create_watch().
* gtypes.h: Move GTimeVal here.
* gthread.h: Remove gmain.h include to avoid circularity.
* giounix.c: Update for new GMain API.
* giowin32.c: Update for new GMain API. (No check for
proper compilation or working.)
* timeloop.c timeloop-basic.c: A benchmarking program for
the main loop comparing the main loop against a
hand-written (timeloop-basic.c) variant.
* tests/mainloop-test.c: New torture test of mainloop.
* docs/Changes-2.0.txt: Started. Added text about
changes to GMain.
* gmain.c (g_main_add_poll_unlocked): Initial fd->revents
to zero. (#8482, Benjamin Kahn)
Sun Jul 30 10:44:16 2000 Tim Janik <timj@gtk.org>
* gmain.c (g_get_current_time): fix tor's recent changes which
got rid of a required variable in the non-windows path.
2000-07-30 Tor Lillqvist <tml@iki.fi>
Finally, a new and improved IO Channel and condition watch
implementation for Win32. Based on code provided by Craig Setera.
When watching file descriptors, for which there is no select()
like functionality on Win32 that would work on all Win32 platforms
for all types of file descriptors (including anonymous pipes), we
start a new thread that blocks while trying to read from the file
descriptor. When the read returns, a Win32 Event is signalled that
the polling routine eventually notices. Meanwhile, the data being
read is stored in a circular buffer, from where the IO channel's
read() method picks it up.
If the buffer fills up the reading thread has to wait for space
becoming available. For this another Win32 Event is used. The IO
Channel's read() method signals this when it has read some data
out of the buffer.
The separate reader thread(s), and the circular buffer(s) with
associated events mean lots of possibilities for fun parallellism
errors. But it seems to work OK, i.e. GIMP runs.
* gmain.c: Small changes to the Win32 polling function.
(g_main_win32_get_poll_func): New function. Perhaps it would be a
good idea to provide this on all platforms.
* giowin32.c: The bulk of the new implementation.
(g_io_channel_win32_wait_for_condition): New function. To be used
where on Unix one does a select() on the channel's fd, like
libgimp's gimp_extension_process(). Could be provided on all
platforms.
* glib.h: Update documentation for IO Channels on Win32. Remove
the declarations for the as of now obsolete old functions related
to IO Channels for pipes with "wakeup" messages.
* glib.def: Some new functions.
* tests/gio-test.c: New file, to test GIOChannel and main loop.
* tests/Makefile.am
* tests/makefile.mingw.in: Add it.
Wed Jul 26 12:59:31 2000 Tim Janik <timj@gtk.org>
* *.[hc]: applied patch from Andreas Persenius <ndap@swipnet.se> that
updates the license headers to the GNU Lesser General Public License,
as well as updating the copyright year to 2000.
2000-04-17 Sebastian Wilhelmi <wilhelmi@ira.uka.de>
* configure.in, acconfig.h: Add configure test for garbage
collector friendliness for GLib. If enabled, ENABLE_GC_FRIENDLY
will be defined.
* garray.c, ghash.c, glist.c, gmain.c, gmem.c, gnode.c, gqueue.c,
gslist.c, gtree.c: If ENABLE_GC_FRIENDLY is defined, NULLify all
memory released by the user, but cached by GLib. This lets a
garbage collector have a more correct view of the actually used
memory.
2000-03-20 Sebastian Wilhelmi <wilhelmi@ira.uka.de>
* gmain.c (g_main_poll): Warn in case of an error during the call
to poll(2). Closes Bug#7564 as reported by David Helder
<dhelder@umich.edu>.
Wed Mar 1 10:39:39 2000 Tim Janik <timj@gtk.org>
* gslist.c (g_slist_reverse): minor optimization.
* testglib.c (g_node_test): added a couple of tests for
g_node_copy().
* glib.h:
* gnode.c (g_node_copy): new function to copy subtrees,
supplied by dbsears@ix.netcom.com.
changed iterator to walk the children list backwards, so
we get down from O(n^2) to O(n).
* gnode.c (g_node_first_sibling): applied patch from
dbsears@ix.netcom.com to optimize access if node->parent
is present.
* gutils.c (g_get_any_init): backed out HAVE_PW_GECOS check around
assignment of g_real_name, sicne HAVE_PW_GECOS is never defined and
thus breaks the original code.
* merged changes from 1.2.7.
Sat Feb 19 19:43:29 2000 Tim Janik <timj@gtk.org>
* testgmodule.c (main): added test to check that not yet bound symbols
in shared libraries of the main module are retrievable, from David Gero.
Fri Jan 28 11:37:41 2000 Owen Taylor <otaylor@redhat.com>
Bug #4156 - Changes vaguely modelled after Scott Gifford's patch
* gtimer.c (g_timer_elapsed): Never report negative times -
clip times to 0.
* gmain.c (g_timeout_prepare): Guard against unexpected
clock shifts by never setting a timeout of more than
data->interval msecs.
Wed Jan 26 05:24:38 2000 Tim Janik <timj@gtk.org>
* glib.h:
* gmain.c: s/current_time/dispatch_time/ for the dispatch() handlers.
refetch the current time after invocation of poll() to cover up for
the time spent in that function call.
Sat Jul 24 20:11:35 1999 Tim Janik <timj@gtk.org>
* merged GLib 1.3.0 with glib-1.2.3 from Fri Jul 16 22:18:36.
* incorporated proposed cleanups from gtk-devel-list.
* bumped version number to GLib-1.3.1
* glib.h:
* gqueue.c:
* gstring.c:
* glist.c:
removed string tokenisation (we got g_strsplit() and g_strjoin()
already) and readline functions.
s/g_list_delete/g_list_delete_link.
implemented g_slist_delete_link.
removed notion of g_ATEXIT() macro in glib.h, this is an *internal*
macro, g_atexit() is provided for public consumption.
added GTrashStack inline utility functions.
reimplement double eneded queues.
removed GStack implementation, people can use a queue or a (singly)
linked list for this task.
deprecated g_strescape(), we need the SunOS variants here.
* gdate.c: added DEBUG_MSG() macro to wrap old messages.
* *.*: CVS merges.
* upgrade to libtool 1.3.3.
Sat May 8 01:52:29 CDT 1999 Shawn T. Amundson <amundson@gtk.org>
* configure.in
gerror.c
gmain.c
gstrfuncs.c
gutils.c
ltconfig
ltmain.sh
gmodule/Makefile.am
gmodule/gmodule.c
gmodule/gmoduleconf.h.in
gmodule/gmodule-beos.c
gthread/Makefile.am: Port to BeOS by myself and Richard Offer.
i.e. gcc -mno-cygwin on cygwin (a.k.a. mingw32, using egcs-1.1.2).
* README.win32: Updated.
* build-dll makefile.cygwin tests/makefile.cygwin: New files.
* glib.h glib.def glibconfig.h.win32 makefile.msc: Slight updates.
* gmain.c: No need to include <fcntl.h> and <io.h> on Win32.
* gmain.c gutils.c testglib.c tests/string-test.c: Test for
NATIVE_WIN32, not _MSC_VER.
* gmutex.c: Must declare g_thread_functions_for_glib_use as
exported (using the GUTILS_C_VAR macro).
* gutils.c gmodule/libgplugin_[ab].c: LibMain not needed.
* gmodule/gmoduleconf.h.win32: Need underscore with gcc.
* gthread/gthread.c: With gcc on Win32, must use memcpy to assign
value of g_thread_functions_for_glib_use (?).
* makefile.msc tests/makefile.msc: Cosmetics.
* configure.in (glibconfig.h): Remove widechar tests and defines.
(fd_set): Change the grep for `fd_mask' to search for `fd_set'.
* gerror.c (fd_mask): Remove conditional typedef. It is not used
elsewhere in the file.
* gmain.c (fd_mask): Likewise.
Wed Feb 24 00:08:42 CST 1999 Shawn T. Amundson <amundson@gtk.org>
* *.[ch]: inserted additional note to look for ChangeLog and
AUTHORS file for a log of modifications.
1999-02-10 Sebastian Wilhelmi <wilhelmi@ira.uka.de>
* glib.h: s/G_LOCK_DECLARE/G_LOCK_DEFINE/ throuhout glib. Added
G_LOCK_EXTERN macro to declare a lock externally.
Mon Feb 1 19:04:28 1999 Owen Taylor <otaylor@redhat.com>
* gmain.c (g_main_iterate): Release the main_loop
lock around calls to prepare() and check() so
that we are not holding the main loop lock
over user code.
Sun Jan 17 14:13:52 1999 Tim Janik <timj@gtk.org>
* glib.h: added a define G_HOOK_DEFERRED_DESTROY, to substitute a noop
GHookList.hook_destroy function.
* ghook.c (g_hook_destroy_link): don't really call hook_destroy if it
is G_HOOK_DEFERRED_DESTROY. for the case where we invoke hook->destroy()
we now clean up the hook completely afterwards, i.e. data, func and
destroy are immediately set to NULL and hook_free can't play with that
values anymore.
* gmain.c (g_source_add): set hook_destroy to G_HOOK_DEFERRED_DESTROY,
instead of using an ugly _noop() hack, this is to avoid an uneccessary
function invokation. set hook_free to g_source_destroy_func, this way
we always invoke the destroy notifiers for user_data and source_data
after execution of dispatch(). thus, g_source_destroy_func() will always
be called within the main_loop lock (this wasn't really assured
before), and can release and reaquire the look around destroy notifier
invokation.
Sun Jan 17 05:12:17 1999 Tim Janik <timj@gtk.org>
* configure.in: incremented version number to 1.1.13, bin age 0,
interface age 0.
* glib.h: added hook_destroy member to struct _GHookList.
* ghook.c (g_hook_destroy_link): if the hook_list defines a hook_destroy
function, use that to marshal hook destruction notifiers.
* gmain.c: removed g_source_free_func and added g_source_destroy_func,
which marshalls the user_data and source_data destructors of sources
outside of the main_loop lock.
removed GIdleData struct since its single member callback can be passed
as source_data directly.
added a gboolean return value to all g_source_remove* functions,
indicating whether the source could be removed, because these functions
don't issue warnings upon unseccessful removals themselves. this way at
least the caller gets a chance to warn upon failing removals.
(g_main_iterate): set in_check_or_prepare around calls to check() or
prepare().
(g_main_pending): simply return FALSE if called from within check() or
prepare().
(g_main_iteration): issue a warning if called from within check() or
prepare() and bail out with FALSE.
(g_main_run): likewise.
(g_source_remove_by_funcs_user_data): new function to remove sources by
user data and function table.
(g_idle_remove_by_data): new function to really remove idles only, since
g_source_remove_by_user_data would remove timeouts or other sources as
well.
* README.win32: More text.
* config.h.win32 glibconfig.h.win32: Update to match the
corresponding generated files on Unix.
* makefile.msc: Update with new source files, and gthread
library. Use the compiler flag -MD instead of using -D_DLL and
"/nodefaultlib:libc msvcrt.lib" in the link phase.
* glib.def: Include new functions, drop removed ones.
* glib.h: Add comments about main loop and polling on Win32. (In
general, it's only for the GIMP's use.) Add Win32 IO Channel
functions. Remove the obsoleted old IO Channel stuff (which was
in #if 0 already).
* giowin32.c: New file.
* gmain.c: Include config.h, conditionalize <sys/time.h>
inclusion. Add g_poll implementation for Win32 (only for the
GIMP's needs for now, it's hard or even impossible to be as clean
and generic as on Unix). Implement g_get_current_time on Win32. If
threads aren't supported, don't try to wake up main thread's
loop. On Win32, use a semaphore and not a pipe to wake up the main
loop.
* gmessages.c: On Win32, allocate a console window if the standard
output handle is invalid before writing to stdout, and reopen stdout
to that console window.
* giochannel.c: Conditionalize unistd.h inclusion. Some indentation
cleanup.
* gstrfuncs.c: Include <signal.h>.
* gutils.c: On Win32, also check the HOMEDRIVE and HOMEPATH
environment variables.
* gmodule-dl.c gmodule-dld.c: In
_g_module_build_path, don't add the "lib" prefix and
".so" or ".sl" suffix if already there.
* gmodule-win32.c: Likewise for the ".dll" suffix.
* gthread-posix.c: Conditionalize <sys/time.h> inclusion.
Thu Jan 7 15:14:08 1999 Owen Taylor <otaylor@redhat.com>
* gmain.c (g_source_free_func): Call the source-specific
free function when the hook is freed not when it
is destroyed; this fixes a bug where a timeout destroyed
from itself would access already freed data.
Mon Jan 4 20:58:50 1999 Tim Janik <timj@gtk.org>
* gscanner.c: eliminated extraneous "register" qualifiers in variable
declarations.
* gmain.c: #undef events and revents which may have been defined in
sys/poll.h for SVR3,4 compatibility on some AIX systems. fix been
provided by Philippe Defert <Philippe.Defert@cern.ch>.
Sat Jan 2 02:20:59 1999 Tim Janik <timj@gtk.org>
* ghook.c:
(g_hook_list_invoke):
(g_hook_list_invoke_check):
(g_hook_list_marshal_check):
(g_hook_list_marshal): avoid unneccessary extra hook referencing (the
explicit hook referencing became unneccessarry with my changes from
Mon Dec 21 21:48:29 1998).
* gmain.c (g_main_iterate): fixed reference counting leaks with
premature loop aborts.
Fri Jan 1 22:47:44 1999 Tim Janik <timj@gtk.org>
* gscanner.c (g_scanner_unexp_token): handle G_TOKEN_IDENTIFIER_NULL
as G_TOKEN_IDENTIFIER.
Wed Dec 23 04:18:11 1998 George Lebl <jirka@5z.com>
* gmain.c: (g_get_current_time) don't cast to timeval since
timeval is for some reason not always a struct of longs, weird
Mon Dec 21 21:48:29 1998 Tim Janik <timj@gtk.org>
* glib.h:
* gmain.c: there was a reference count race for hooks during invocation
loops. since all (known) hook loop implementations, do currently start
out with g_hook_first_valid() and iterate with g_hook_next_valid(),
g_hook_first_valid() will now return a referenced hook, and
g_hook_next_valid() will "eat" that, and eventually transfer it to
the next hook. <sigh> unfortunately this requires g_hook_next_valid()
to take the hook_list as additional argument.
* gmain.c (g_main_iterate): adjusted callers of g_hook_next_valid().
Mon Dec 21 03:48:04 1998 Tim Janik <timj@gtk.org>
* gmain.c (g_main_iterate): default initialize source_timeout with -1
so we have a sane timeout value if (*prpare) doesn't set it.
Sat Dec 19 16:56:02 1998 Owen Taylor <otaylor@redhat.com>
* gmain.c (g_main_poll): Allocate space for pollfd's
_after_ adding poll wake-up-pipe record.
* gmain.c (g_main_add_poll): Changed name
of internal function g_main_add_poll_unlocked()
back from the non-sensical g_main_add_unlocking_poll().
Fri Dec 18 00:03:17 1998 Tim Janik <timj@gtk.org>
* glib.h:
* gmain.c:
(g_main_is_running): new function to check whether a main loop has been
quitted.
(g_main_new): added a gboolean argument to determine whether the loop
should be considered initially running or not. however, g_main_run ()
will still reset the main loops running state to TRUE upon initial
entrance.
* gmain.c:
(g_main_iterate): documented this function's purpose in 5 steps.
for step 2), flag sources as G_SOURCE_READY even if !dispatch and
check G_SOURCE_READY prior to (*prepare), so we don't call (*prepare)
on them multiple times.
Thu Dec 17 23:43:47 1998 Tim Janik <timj@gtk.org>
* gmain.c (g_main_add_poll): reordered arguments, so GPollFD* comes
first, <sigh> (sorry Snorfle, i should have let you know in the first
place).
(g_main_dispatch): stack G_HOOK_FLAG_IN_CALL flags. call source's
destructor when destroying a source.
Wed Dec 16 22:32:13 CST 1998 Shawn T. Amundson <amundson@gtk.org>
* glib.h:
giounix.c: s/g_main_poll_add/g_main_remove_add/g
s/g_main_poll_remove/g_main_remove_poll/g
s/g_main_poll_add_unlocking/g_main_add_unlocking_poll/g
(from Tim Janik)
* gthread-posix.c: use g_free in mutex_free (from Tim Janik)
Thu Dec 17 04:10:49 1998 Tim Janik <timj@gtk.org>
* glib.h (G_LOCK_DECLARE_*): if !G_THREADS_ENABLED, eat the
trailing semicolon with a bogus function declaration, instead
of with a bogus variable declarations, so we avoid unused
variable warnings.
1998-12-16 Sebastian Wilhelmi <wilhelmi@ira.uka.de>
* docs/glib-config.1: Updated to reflect the existence of gthread.
* gmain.c (g_main_poll_add_unlocking): Changed
g_main_poll_add_unlocked to g_main_poll_add_unlocking to match
semantic, (indeed, main_loop must be locked, when calling this
function). Removed the unlocking from the end of that function, as
that is not right. Made a 'HOLDS' comment above the function.
Wed Dec 16 03:16:58 1998 Tim Janik <timj@gtk.org>
* configure.in: version bump to 1.1.8, binary age 0, interface age 0.
* glib.h: changed g_lock() to G_LOCK(), g_unlock() to G_UNLOCK() and
g_trylock() to G_TRYLOCK(), since these are macros that expand to
nothing with --disable-threads.
changed G_LOCK_DEFINE() to G_LOCK_DECLARE() and introduced
G_LOCK_DECLARE_STATIC() to achive the results of static G_LOCK_DECLARE().
changed semantics of g_thread_supported to g_thread_supported() so it
can be used as a function like g_module_supported(). the actuall
definition is still a macro that expands into a variable for
performance reasons though.
various indentation and coding style cleanups.
* configure.in: added --enable-threads that defaults to yes.
* gmutex.c: changed tests g_thread_supported to g_thread_supported (),
changed variable settings of g_thread_supported
to g_threads_got_initialized.
garray.c:
gcache.c:
gdataset.c:
gdate.c:
ghash.c:
glist.c:
gmain.c:
gnode.c:
gslist.c:
gstring.c:
gtree.c:
gutils.c:
changed s/g_lock/G_LOCK/, s/g_unlock/G_UNLOCK/,
s/static G_LOCK_DEFINE/G_LOCK_DECLARE_STATIC/.
