Original commit message from CVS:
rewrite vararg prop entry using functions to cope with GST_PROPS_TYPE_LIST - fixes loads of segfaults in caps rewrite
Original commit message from CVS:
fix threading on non-optimized atomic architectures. This unfortunately requires initializing threading in init_pre because the debugging subsystem needs this. Therefore the --gst-nothreads command line option does not work anymore. So I erased it. You can however still call gst_use_threads (FALSE); before calling gst_init
Original commit message from CVS:
GST_DEBUG reorganization
This is a big diff (ca 450k), containing loads of stuff:
- gstinfo.[ch] complete rewrite
- changing of all GST_DEBUG messages to reflect that change
- reorganization of subsystem disabling
- addition of gstconfig.h.in so we can track the disablings
- <gst/gst.h> does not include <unistd.h> and <config.h> anymore
- documentation updated for gstinfo stuff (build the docs yourself to know what changed)
- bugfixes for making of the docs (files from CVS are not deleted anymore
- testsuite for debugging changes in testsuite/debug
expect breakage
Original commit message from CVS:
- further productizing the fixed flag on GstCaps by using the flag on
GstProps instead of keeping the two in sync.
- implemented proper flag updates in _props_set and _props_remove.
Original commit message from CVS:
- added GST_ELEMENT_ERROR flag to mark elements in error and to trap and
avoid recursive error loops.
- moved state change signal back to original place as it causes lockups.
Original commit message from CVS:
I don't know how I ever got to introducing this madness, but I was wrong. *REALLY* wrong. It hasn't even worked, it still doesn't work and it will never work because the whole concept of doing a multifile output in filesink is completely braindead. Signal EOS, close file, set new filename and reset to PLAYING is so much easier. This patch removes all the multifile output cruft that I introduced and makes filesink simply work.
Original commit message from CVS:
Add new --enable-fast-stack-trash option, defaults to on.
We need this because valgrind doesn't implement an insn used
by it.
Original commit message from CVS:
Add patch for better buffering (especially useful for network streaming), see #108268 - kudos go to janzen@pixelmetrix.com
Original commit message from CVS:
- the result from the getcaps function should be sunken (gst_caps_sink)
instead of unreffed. Not sure that's the right thing to do...
Original commit message from CVS:
- moved functionality of GST_FORMAT_UNITS to GST_FORMAT_DEFAULT, the old
_DEFAULT functionality was ambigous; now the _DEFAULT format is media
specific (frames for video, samples for audio, ...)
Original commit message from CVS:
- First attempt at implementing #113180, this one also removes
automatic (byte) offset calculations in the subbuffer code and moves
that logic to bytestream and filesrc. It will also update the offset
and duration fields in some special cases.
Original commit message from CVS:
Fix aliasing problems noticed by gcc-3.3. Suggests that we need to
change gst_pad_{push|pull}() to GstData *. (and probably others)
Original commit message from CVS:
Change GST_DEBUG_FUNCPTR() to return a pointer of the same type that it
is called with. Fix lots of warnings due to change.
Original commit message from CVS:
This is filtered caps in 20 lines. Implemented full featured parsing of pipelines with filtered caps. To get a grip of the syntax you might want to look at testsuite/caps/string-conversions.c or run that test
Original commit message from CVS:
added gst_{props,caps}_{from,to}_string to API, see documentation for use, put errors into testsuite/caps/string-conversions.c
Original commit message from CVS:
SysClocks need their own mutex/cond pair, as we have multiple (via subclasses). Use a static mutex for setting THE system clock
Original commit message from CVS:
some minor improvements:
- better debugging output
- don't set state to paused when plugging anymore
- bugfix: set typefind caps on srcpad, not sinkpad
- remove unnecessary check for right caps
Original commit message from CVS:
implement proper refcounting of pad's templates and references from real pads to their ghost pads and the other way around - this fixes the segfaults in the player
Original commit message from CVS:
- Explicitly set buffer fields to default values instead of using memset,
this seems to speed up buffer allocation by at least 30%
Original commit message from CVS:
- remove some \n in GstThread debugging
- move the state change g_cond_signal to the set_state function to make
sure that we don't unlock the waiters before we completed the state
change.
Original commit message from CVS:
I suck - it's the wrong variable and I don't get it for ages even while testing this code
But while I was atit, I fixed it right this time (I hope)
Original commit message from CVS:
* make the user registry location major-minor versioned so it doesn't
clash with older incompatible registries
* make gst-register output the location of the registry it's writing
Original commit message from CVS:
assorted fixes:
- fix for #111146 (works now with lots of warnings (or maybe even without - the fun of races))
- more useful debugging output
- restructuring of when to release the lock
- emitting SHUTDOWN when holding lock so we're not destroyed while signalling
- probably something else I don't remember
Original commit message from CVS:
- removed an assert, mmaping the stack of another thread seems to work fine.
This makes the cothread based scheduler work again with the new GstThread.
Original commit message from CVS:
- make child_state_change a virtual function
- remove unneeded != NULL checks
- implement gst_bin_sync_children_state
Original commit message from CVS:
Added a simple function gst_pad_recover_caps_error that a plugin
should call from now on (instead of gst_element_error). This function
will emit a caps_nego_failed signal that the app can catch to resolve
the negotiation failure.
Original commit message from CVS:
It's absolute fun if you don't find this. If an element claims a successful state change but did not actually change its state it is broken and the program deserves to crash. And hard. Can you hear I'm annoyed?
Original commit message from CVS:
Ouch: gst_element_dispose does a state change to NULL. Change the state to NULL yourself if you dispose plugins that reference some of the freed memory on state changes.
This is probably broken in lots of cases.
Original commit message from CVS:
API change: rename (currently unused function) gst_(data|buffer)_needs_copy_on_write to gst_(data|buffer)_is_readonly - we don't want to confuse even core developers
Original commit message from CVS:
- Added infrastructure to run a filter against plugins, registries and
registrypools. This makes it possible to create custom code to filter
out the plugins/features you're interested in.
Original commit message from CVS:
enable debugging output for bison parser, fix // comments, better error recovery, error out on non-instantiable elements (fixes#110758)
Original commit message from CVS:
Revert last Makefile.am change, and fix it correctly. I _think_
this won't break with earlier flex versions, otherwise let me know.
Original commit message from CVS:
- added macros GST_CLOCK_TIME_IS_VALID and GST_BUFFE_TIMESTAMP_IS_VALID
- use macros in some places
- buffer's timestamp field is a GstClockTime, not a guint64 - this is currently the same, but be sure to only use GstClockTime when working with timestamps in the future
Original commit message from CVS:
general fixes:
- changed newly added locked_state API to work like GStreamer does
- added gst_element_sync_state_with_parent function
- revert: pad linking does not require non-playing state
- updated spider and pipeline parsing to link elements in playing state
- bugfix: pads are now activated when added to a playing element (broke above change)
Original commit message from CVS:
rewritten pipeline parsing lands. Have fun breaking it.
regressions:
- No support for filtered links. If anybody needs this, please contact me and I'll add it as fast as possible.
improvements:
- The pipeline building actually works as expected.
- syntax compatible nearly everywhere but more flexible
- better property parsing (you can now set enums by name or nick)
- uses locked_state to allow for delayed pads
- can connect video pipelines without the need for queues
- allows properties on bins
- does not return a bin, but an element.
You may want to read docs/random/company/gstparse to learn how it works.
Original commit message from CVS:
huge modifications, hope i didn't break something:
- added functions for setting/getting locked state of elements. gst_element_lock_state syncs the pads state with the state of its parent.
- made all gst_element_link_* functions call gst_element_link_pads_filtered
- gst_element_link_pads_filtered now accepts NULL as a pad name.
- make sure the link functions are only called when the elements are paused
That's it.
