Original commit message from CVS:
A rather large patch:
- changed the API for the padtemplates:
- remove the factories (array of pointers) for the padtemplates,
properties and caps. The static array was a nice idea but converting
all the property values to a gpointer was not a good idea.
float properties were not possible, and casting a gint to a pointer
is not very portable. The new API just uses the _padtemplate_new,
_caps_new and _props_new functions to create the templates.
This has the added benefit that the API is now uniform for static
and dynamic templates and that the code can be made cleaner.
- lots of cleanups in the way the capabilities are constructed (va_list)
- lots of updates for all the plugins (new API)
- docs updates (new API)
- removed the videoraw docs.
Original commit message from CVS:
Lots of updates to the plugins for caps negotiation.
Added YUY2 output to the win32 dlls.
Added a colorspace converter in gstplay
Original commit message from CVS:
Add a proposed makefile to use in place of Makefile.am.
This depends on automake 1.4d or later, but is much neater, avoiding
horrible hacks to pass special flags in.
I put it here for safekeeping until automake 1.5 is released, at which
point we can reasonably require it.
Original commit message from CVS:
Fix obscure segfault I just got: perhaps due to running with libxml 1.8.11
(prerelease version), but a worthwhile fix anyway.
Original commit message from CVS:
A rather large update:
- a new videosink using plain X windows (xvideosink). It uses capsnego to
set up it's formats.
- gstplay uses the new xvideosink and a gtk_socket. The initial window
size is set to 0x0 so you need to resize it before you see something.
- got rid of the video metadata include file in favour of properties and
caps negotiation
- adjusted most plugins to not use the metadata.
Original commit message from CVS:
Simplified the FOURCC API to the caps, add a GST_MAKE_FOURCC macro
to create a fourcc.
Adjusted the plugins to use the new FOURCC API
Original commit message from CVS:
Added a plain X videosink, the videosink uses capsnego.
adjusted the v4lsrc so that it uses capsnego on the src pad
Small fixed to capsnego and the properties/caps.
Original commit message from CVS:
Modified a lot of plugins to use the caps system.
Modified the caps of audio/raw to our agreed properties.
Added the multidisksrc plugin of Dominic Ludlam
Renamed audiosink/src to osssink/src and updated all the examples using
the old name. Moved oss specific plugins in an oss directory. removed
the old audiosink from the elements/ dir.
removed audioraw.h metadata header files since we now use the properties.
There are still a few plugins that won't build because they include the
old audioraw.h header file. This will be fixed soon.
Make sure the caps are set in the plugins as described by their
padtemplates (this should solve problems with gstmediaplay with various
media files).
*please don't panic when some plugins won't build, just cd manually into
the plugin dirs* This will be fixed soon.
Original commit message from CVS:
More work on capsnego proxying. It should be OK now.
Added another testcase enum that shows various capsnego algorithms.
Warn about pads that try to set a capability incompatible with their
padtemplate.
Implemented refcounting and copy_on_write for caps/props.
Original commit message from CVS:
More work on capsnego, mostly proxying
Added another testsuite for capsnego
Added caps to vorbisdec, mp3parse, mp1videoparse
Redid the queue proxy handling a bit.
Original commit message from CVS:
Reworked the capsnegotiation function
audiosink now uses capsnego to set its parameters
mpg123/ac3dec use capsnego instead of metadata
Added the beginnings of a testsuite for capsnego.
Original commit message from CVS:
Fix gst_init() so that it doesn't fail when passed two NULLs as parameters.
Was dereferencing the argv pointer to get progname: now defaults to
"gstprog".
Original commit message from CVS:
Small fixes to pipeline and gstbin.
Fixed the xmmstest makefile entry
Removed the plugin_load statements in videotest
Applied the patch from Steve Baker for float args parsing
Original commit message from CVS:
Added a utility function in gstutils to set an object argument as a
string. gstparse.c and gstelement.c now use this function.
Original commit message from CVS:
CPU detection of MMXEXT and probably 3DNOW (not on Athlon yet)
Renamed the mmxe motion compentation files.
disabled SSE optimisation in ac3dec because it apparently faults..
enable MMXEXT motion compensation on non SSE CPUs (AMD)
Original commit message from CVS:
slightly changed the Makefile.am in videoscale libs.
v4lsrc works again
A little modification to the videotest programm to use a tee element.
added support for multiple paths in --gst-plugin_path=.
added ab ENV variable GST_PLUGIN_PATH to specify extra directories to
look for plugins.
Original commit message from CVS:
Added two options: --gst-plugin-spew will print errors when plugins fail
to load. --gst-plugin-path= will add a directory to the path searched for
plugins.
Original commit message from CVS:
More doc updates
Manual additions: xml/load save of custom XML, debugging, request pads
added a signal to the example plugin
fixed a bug in mp3parse (ifdef 1)
Original commit message from CVS:
Fix a couple of instances of childs, which should be xmlChildrenNode. We
need to be careful to stop these creeping back in... ;-)
Original commit message from CVS:
- backported the xml load bug in the properties
- backported the videosink bug
- make sure the regiontype is set to NONE after a getregion
- an ugly fix (copy the chainfunction outselves) to make the avidecoder
work again.
Original commit message from CVS:
Reworked getregion/pullregion stuff. Region is now specified by a type
(OFFSET_LEN, TIME_LEN, etc.) and two guint64's. They are offset and len,
where offset can be used for the time field, and len would be zero or
the time unit (say, 33ms for video).
Original commit message from CVS:
Added a first attempt for XML embedding in the main core XML.
Objects will emit a signal that a user app can connect to in order to
insert its XML into the tree.
You can catch the object_loaded signal in GstXML to parse the user
supplied XML data in the stream. The object_loaded signal is implemented
with a custom made class signal. All GstObject classes now automatically
create a GstSignalObject that serves as a proxy to the user app when an
object is loaded. All objects are currently responsible to emit the
class signal themselves.
runxml and createxml serve as an example how the XML hooks can be used to
insert and retrieve custom XML tags.
Original commit message from CVS:
Lots of plugins fixes where the parentage macros were replaced with the
correct gst_*_get_parent.
Updated the example.
'gstreamer' now is the doc type instead of GST-Pipeline
Original commit message from CVS:
This is a megapatch with the following changes:
- moved the gchar *name to GstObject, removed the ones in GstElement and
GstPad.
- moved the parent handling completely into GstObject. This cause *all* of
the plugins to fail (except those that used gst_pad_get_parent)
- rearanged the XML save handling.
- GstObject now has a class function save/restore_thyself.
- GstObject has a generic method gst_object_save_thyself, this makes it
possible to fire a signal wehever a new object is loaded. This is needed
so we can add XML save hooks.
- GstXML API has changed slightly. You now have to create a GstXML object
first before you can actually load something. This makes it possible to
attach a signal to GstXML whenever an object is loaded. I'm not sure we
will keep this interface.
- GstObject can now print the path_string without knowing about the GstPad and
GstElement types.
- Added gst_bin_get_by_name_recurse_up to lookup an element in the current
element hierarchy.
- added XML namespaces to the saved pipelines the namespace is:
http://gstreamer.net/gst-core/1.0/
namespaces are needed to distinguish user generated XML from the core XML.
Note that the plugins still contain a macro GST_OBJECT_PARENT that will be
replaced with gst_pad_get_parent shortly.
Original commit message from CVS:
Add gst_element_request_compatible_pad and remove gst_element_request_pad.
Implemented something reasonable for gst_element_request_compatible_pad,
but havn't tested much: it won't work for tee because the pad templates
have no caps, and negotiation is not yet written, so it is assumed that the
tee pads can't connect to anything.
Original commit message from CVS:
Docs updates
Added XML load from memory functionality
Undid the videosink patch, something else is wrong now on my machine:
no MMX acceleration :-(
Original commit message from CVS:
Applied the patch from Sepp Wijnands <mrrazz@garbage-coderz.net> that solves
the 'OSS hangs when opening the device when in use' bug.
Original commit message from CVS:
Check for NULL buffers before chaining/pushing them to the next element.
NULL buffers happen on EOS.
gst_bin_iterate now returns a gboolean indicating something usefull happened.
It the bin is EOS, it returns FALSE.
Original commit message from CVS:
More EOS changes.
When a bin2 is found inside a bin1, we add the bin2 to the EOS providers
of the bin1. When there is nothing more to schedule in bin1 and bin2 has
fired EOS, bin1 is in EOS.
The queue overrides the EOS notification and calls EOS on the src pad
when the queue is empty and the sink pad is in EOS.
Original commit message from CVS:
A very small change to make eos somewhat work. no inner bins are checked.
When an element fires EOS, the chain with that element is removed from
the scheduler (marked inactive). If all chains are inactive, the bin
fires EOS.
Original commit message from CVS:
Major cleanup of the latest ghostpad changes. Fixed everything that
broke, correctly. Someone will want to go update the API doc templates.
Original commit message from CVS:
ALPHA COTHREADS WORK! Worked around a nasty stack issue that probably
can't be solved anyway. Tomorrow the UDB build will commence, and let the
best guess win!
Original commit message from CVS:
First pass at updating to new ghostpad system. The objects are in place,
I now need to go and get all the Bin end of things worked out. Testing
should be fairly easy, at least for verification.
Everything I've tried so far works with no changes, with is amazing.
That's just cool. Once again we rewrite an entire subsystem, and nothing
else notices anything but the new features ;-)
Original commit message from CVS:
Added mthodes to request an element to create pads: gst_element_request_pad*
This can be used to construct a tee and a muxer/mixer/aggregator element.
Moved the tee element to elements/ because it can now be handled with the
new pad request features.
The padfactory also has some changes: a pad can now be of presence REQUEST,
which means that the pad can be requested from this plugin (doh).
Original commit message from CVS:
Updates to cothreads code, including non-working alpha. Changed things a
bit, including PPC. Not having a PPC machine, I need someone to test
these changes and report back whether they worked or not.
Original commit message from CVS:
Adding nasty hack to rules to generate cothreads.{o,lo}, to get dependencies
right. Rules copied from automake, and therefore a bit dependent on automake
keeping doing dependency things the same kind of way, but it should work as
long as automake puts dependencies into .deps/*.P
Original commit message from CVS:
Massive build fixup. Will send message to -devel list later with details
on the changes and what they mean for Makefile.am writers. Check
docs/random/omega/build/TODO for a list of things that I had to make sure
of.
NOTE: this requires a complete rebuild of all plugins, since I also
changed the STATE enum to a bitfield instead of sequential numbers.
Original commit message from CVS:
Added an extra signal_cond to queue to make sure that the waiting thread
is woken up. Can somebody with queue problems verifify that this does
improve the situation a bit. I'm suspecting that something else is going
on, like a pthreads bug or something.
Small updates to the fake elements.
Original commit message from CVS:
Handle the tee element like bin, thread, pipeline, so that it can be used
in gstreamer-inspect and gstreamer-launch...
Original commit message from CVS:
gstreamer-register will now never cause the
** WARNING **: gstplugin: registry needs rebuild
warning to appear - sets a hidden global in gstplugin.c to turn the
warning off.
Fix a warning in gst/cothreads.c (function was declared extern, defined
static)
Original commit message from CVS:
Check plugin repository up-to-dateness by scanning through all directories
on path, and checking that they and all their files are older than the
repository. Give warning message if not.
Original commit message from CVS:
Config directory (which currently just holds the registry)ecan now be set
by configure --with-configdir=<directory path>
Original commit message from CVS:
Updated the manual and the docs.
Removed the esdsink in gst/elements/ we have a real one not in the
plugins dir.
Added more APIs to query the plugins, types and caps. more fields now
have a getter and a setter. This is needed to make gstreamer wrapper
fiendly.
Added gst_element_disconnect beacuse we also have a gst_element_connect
Original commit message from CVS:
Rearranged cothreads sources a bit, added some API docs.
Added some functions needed for gstreamer-inspect.
Added num_sources to fakesink.
Original commit message from CVS:
Added a check for libjpeg
Added and action for the load menu in gstplay
Fixed drag and drop in gstplay
More verbose errors in gstasyndisksrc
Original commit message from CVS:
Massive, massive update of most source files. I went through and cleaned
up all the warnings that I could, which involved fixing some of the plugins.
The configure.in script was re-arranged and cleaned up so, and a check for
libtool 1.3.5 was added to autogen.sh. Added checks for Gtk and GNOME.
Some plugins were removed from the list of things to build for various reasons.
Added GST_DEBUG_FORCE_DISABLE in gstgetbits, since that's time critical and
even an if() from a DEBUG can significantly increase runtimes.
Original commit message from CVS:
Mega update of INFO, DEBUG, and ERROR subsystems, renamed with GST_ prefix.
GST_DEBUG now takes a category parameter, which is the same as GST_INFO
system. They are now called GST_CAT_*. All the GST_DEBUGs are set to 0
for now, we need to go and fix all these eventually.
Original commit message from CVS:
Added --enable-info and --disable-info-system
nothing specified = INFO enabled, mask to 0x00000001
--enable-info = INFO enabled, mask to 0xffffffff
--disable-info = INFO compiled into a nop and thus disabled
Original commit message from CVS:
Cleaned up INFO system some more, added API to select categories, modified
the printout function to print cleanly for GST_INIT info (the default min)
and spew function:line debug_string [element] for everything else.
Eventually could make even that configurable easily enough, just check
against another bitmap instead of checking for == GST_INIT.
Original commit message from CVS:
Started work on better ghostpad management, and started to build the low-
level EOS mechanism.
Also removed a couple of printouts that aren't needed any more.
Original commit message from CVS:
Added multiple caps to the pads
Extended the padfactory for the multiple caps
GstCaps now have a name so that future manipulations on them can be done by name
Updated the plugins for the new caps list
Original commit message from CVS:
Fixed autoplugging for avi
Fixed a bug in multiple typefind function loading: when a typefind is
requested for a mimetype, all plugins with a typefind function should be
loaded.
Original commit message from CVS:
Docs updates: remove sink, src, filter, connection
added gstinfo to docs.
Updated some core files for the docs
gsttypefind did not include gst_private
Added CFlags to Makefile.am in gst/elements
Original commit message from CVS:
Fixed buffer flag handling
gstplay used old flag handling
updated some plugins for the new objects/error handling
Fixed a serious buffer error in gst_buffer_append
Original commit message from CVS:
Fixed a serious bug in gst_props_new: properties with a 0 value causes
a segfault
Fixed a possible bug in gstautoplug when one of the constructed paths
was empty
Commented out a too liberal assert in gstscheduler.c
Added GST_BUFFER_DISCONTINUOUS
tried to fix asyndisksrc
Original commit message from CVS:
Added DEBUG, INFO, and ERROR systems. Very little code is converted yet.
Policy decisions need to be made as to what kinds of cases get what kind
of response, and what the default ERROR response should be. Right now it
will print out all the information, then have gdb spew a stack trace.
Original commit message from CVS:
Reparented everything to GstElement, removing GstSrc, GstSink, GstFilter,
and GstConnection. Also fixed a bug in gst_bin_iterate_func.
Original commit message from CVS:
Pending editor changes.
fixes for autoplug of elementary MPEG1 video streams
make sure mp3 types get compiled
videosink changes for xml loading
Original commit message from CVS:
Mostly minor little changes, but two interesting things:
1) removed a pthread_join from the gst_thread_main_loop. commented out
because the thread isn't supposed to run pthread_join, the main process is.
2) Fixed a major bug with cothreads in threads. Had to add MAP_FIXED to
the mmap() of the cothread stack. Presumably the gilbc that ships with
redhat 7.0 now places these mmap requests somewhat randomly. Since they
*must* be exactly where we expect them, it was failing. MAP_FIXED forces
it to put it where we say.
Original commit message from CVS:
Added code to attempt the scheduling of chained cases. Very simply right
now. Also added code to allow the use_cothread flag of a bin to force all
the chains to use cothreads as well.
Original commit message from CVS:
More massive changes to the scheduling system. Moved the scheduling code
to gstscheduler.[ch], so an child bin can replace the scheduler.
Introduced the concept of chains, which are subsets of the list of managed
elements for a given manager bin, which get scheduled as separate entities.
gst_bin_iterate_func should be pretty much fixed now, the scheduling code
gets to do all the hard work.
Cothreaded case work in the couple tests I've tried, chained is next.
Original commit message from CVS:
Lots of editor changes:
- restucturing of object creation
- loading of xml files
- visual eye candy: color changes on mouseover
- active object is indicated
- property box changes for boolean and enum types
- property box: create properties per element
- show pads/caps in propertybox
added gst_util_get_bool_arg in gstutils.c
added default properties for audiosink
Original commit message from CVS:
Added handoff signals to fakesrc and fakesink
Added scheduling property to identity
Added GST_STATE_TRANSITION macro to check for state changes.
Modified gstbin and gstthread to the new state change macros
Original commit message from CVS:
Fixed a bug in gstbin.c the manager was not detected.
Some other testcases
bin, thread, pipeline use the elementfactory on gst_*_new
Original commit message from CVS:
Implemented some test cases... they all fail :-(
Added a flag to identity to for loop_based behaviour.
Added a signal to fakesink when a buffer is consumed.
Original commit message from CVS:
Fixing two bugs:
1) I had commented out an unlock. duh.
2) changed the _get function to return the buffer rather than call _push
Also uncommented some debugging I'd turned off. Need to solve the verbosity
problem somehow, I think by way of debug levels as well as info levels...
Original commit message from CVS:
Massive scheduling changes (again). Not entirely complete, but getting
closer. Need to think about various scheduling plans that we might want
to produce, and figure out the rules for what is legal, and how to get
the results we need as far as the plan.
Original commit message from CVS:
Added API to dynamically create GstCaps and GstProps
Changed typefind of mpeg1 to set the GstProps.
autoplugging now works again for mpeg1, whoohoo!
Original commit message from CVS:
Cleanup in gsttypes.c:
removed the crazy GList of GHashTables, since the autoplugger will
use the GstCaps and elementfactories instead of the type system.
We don't maintain a list anymore of the elements for the specific
removed unused code in XML loading/saving.
Cleanup in gstelementfactory:
removed the register/unregister methods, register is now implicit when
gst_elementfactory_new is called. _unregister is now _destroy.
Removed logic to register/unregister the types in gsttypes.
added methods to query if the factory can src/sink a GstCaps
Make sure the elementfactory is set in the element_class when a new
element is registered with gst_elementfactory_new.
gst.c: properly register the basic bins
gst_pipeline: use new gstautoplug (next checkin)
gstprops: fixed an error in compatibility check
registry test program changes
plugins: misc changes for the new caps system.
Original commit message from CVS:
The typefind function now returns a GstCaps structure instead of a gboolean.
modified some plugins to this new behaviour
Fixed autoplugging for the mpg123 case. When an element is selected in
autoplugging and the caps do not match, another element is selected until the
caps match. only examples/autoplug works because gstmediaplay uses a threaded
setup that does not seem to work with the current scheduling.
Original commit message from CVS:
some editor changes= loading of save pipelines (not working)
added include in gstdebug.h for getpid
property loading implemented
GstXML can now be queried for the toplevel elements.
Original commit message from CVS:
WARNING: Don't grab this updated unless you're really, REALLY sure.
WARNING: Wait for the next one.
Whole lotta changes here, including a few random bits:
examples/*/Makefile: updated to use `libtool gcc`, not just `gcc`
gst/
gstbuffer.h: updated to new flag style
gst.c, gstdebug.h: added new debugging for function ptrs
gstpipeline.c: set type of parent_class to the class, not the object
gstthread.c: ditto
plugins/
cdparanoia/cdparanoia.c: added an argument type, updated some defaults
cobin/spindentity.c: updated to new do/while loopfunction style
mp3encode/lame/gstlame.c: argument types, whole lotta lame options
tests/: various changes
Now, for the big changes: Once again, the scheduling system has changed.
And once again, it broke a whole bunch of things. The gist of the change
is that there is now a function pointer for gst_pad_push and gst_pad_pull,
instead of a hard-wired function. Well, currently they are functions, but
that's for debugging purposes only, they just call the function pointer
after spewing lots of DEBUG().
This changed the GstPad structure a bit, and the GstPad API as well.
Where elements used to provide chain() and pull() functions, they provide
chain() and get() functions. gst_pad_set_pull[region]_function has been
changed to get_pad_set_get[region]_function. This means all the elements
out there that used to have pull functions need to be updated. The calls
to that function have been changed in the normal elements, but the names
of the functions passed is still _pull[region](), which is an aesthetic
issue more than anything.
As for what doesn't work yet, just about anything dealing with Connections
is hosed, meaning threaded stuff won't work. This will be fixed about 12
hours from now, after I've slept, etc. The simplefake.c test works in
both cothreaded and chained cases, but not much else will work due to the
Connection problem. Needless to say, don't grab this unless you *need*
these features *now*, else wait to update this stuff until tomorrow.
I'm going to sleep now.
Original commit message from CVS:
Move a step closer to autoplugging. The element selection in the
pipeline works.
Slight modification to the typefind function handling.
removed the stupid mime types in the mpeg decoders.
Adjust videosink for the new caps handling.
Modified the queue example to show an error in plan creation; Erik can you
solve this? This is the type of pipelines autoplugging uses.
Original commit message from CVS:
Header cleanup: try to include as little as possible; this will probably
speed up compilation a bit.
changed the .c files to use #include "..."
Fix for the 'plugins are loaded twice' bug.
Fix 22186: GstObject flags are now used everywhere. Added *_FLAG_LAST so
elements do not use the same flags. Added some padding in the flag enum
for future expansion.
Original commit message from CVS:
Changed the GstPadFactory and added the GstPadTemplate.
The GstPadFactory is an easy way to define a pad with capabilities.
The GstPadFactory is converted into a GstPadTemplate. The template is
used to create new pads and to expose the possible pads used in an
element to the plugin system.
updated mp3parse, mpg123, cdparanoia to the new API.
Rerun gstreamer-register because the XML definition has changed.
Original commit message from CVS:
More Caps changes, this change introduces the padfactory. the GstPadFactory
will be used to describe the pads that an element will contain during
its lifetime.
mpg123 and mp3parse serve as an example.
More XML changes to incorporate the padfactories.
Original commit message from CVS:
Implemented the hybrid scheduling system for sources and connections
outside the current Bin. Is a bit hackish in one place, but I'll work out
a way to make that cleaner soon. queue.c in tests now works nicely in all
cases. More to come later.
Original commit message from CVS:
First attempt at rebuilding the type/plugin system
- make sure caps/props are saved in the registry
autoplugging is completely broken.
removed the typefactories and typeids from all the pads in the plugins
XML load/save is ok (be sure to rerun gstreamer-register)
Original commit message from CVS:
Snapshot of work-in-progress do deal with out-of-bin elements. Current
system is very likely going to be ditched due to its complexity and the
fact that it doesn't work right now. More than that, it's generating some
really odd results in my test program, which aren't readily explainable.
If you want to keep a working copy of your working copy (heh), don't update
until this is replaced with a hybrid approach. This approach will be the
beginning of a plan generator that can construct full hybrid schedules
given hints from various places.
Hybrid means that cothreads are used, but there are chunks of the pipeline
that are actually dealt with by chaining. This can improve speed by
reducing cothread switches (which are much cheaper than thread switches,
but still far from free, about 570 cycles on my PIII), but does carry a
complexity burden. Luckily, the structure of GStreamer allows that burden
to live entirely in create_plan. Luck? I think not ;-)
Original commit message from CVS:
Another big set of changes. Connections are now also pullfunc based.
gstqueue has been updated, I don't know of any other connections offhand.
There are still a few things that need doing, specifically the concept
of a source or connection with connections to multiple thread contexts is
not dealt with. This may force us to move the threadstate from the
element to the pad, maybe keeping the element's copy for simple cases.
Then the Bin would create a structure to pass to the cothreaded _wrappers
of any such elements, which would detail the pads that are to be dealt with
by this particular cothread context.
That will speed things up to, since we don't have to look through the list
of all pads for every Src or Connection element for every iteration, we can
simply step through the list provided by the plan. Special case might even
have a single pad pointer sitting there to trump the list, if there's only
one (the common case anyway).
Task 23098 is tracking these changes. The main task 22588 depends on that
subtask, as well as 22240, which is a consistency check on PAD_DISABLED.
Original commit message from CVS:
Changed the way things are scheduled, especially sources. A Src used to
have a push() function, and optionally a pushregion() to deal with async
reads, etc. That whole thing has gone away, in favor of providing a
pull() function for the output (Src) pad instead, ala chain functions.
This makes constructing cothreaded schedules out of non-loop elements
somewhat easier. Basically there was always a question as to which pad
was being dealt with. In the pullregion case, cothread-specific data was
used to try to pass the region struct to the right place, which is a slow
hack. And in general, the push function severely limited the kind of
tricks that could be played when there's more than one output pad, such as
a multi-out file reader with async capabilities on each pad independently.
This changes the way cothread scheduling occurs. Instead of the hack to
deal with Src's by calling their push() function (or optionally the
pushregion(), in certain cases), we now are working towards a general
mechanism where pads are the only thing that are dealt with directly.
An optimization was made in the process of doing this: the loopfunction
actually run as the outer [stack] frame of the cothread is now set more
intelligently in create_plan() based on what kind of element it is. We
now have:
loopfunc_wrapper: used for loop-based elements, it simply calls the
loopfunc in a loop, paying attention to COTHREAD_STOPPING (see
below). It currently does other, soon to be depracated, stuff.
pullsrc_wrapper: wraps a Src that's not loop-based (since your options
are now loop- or pull-based)
There will be a couple more to deal with other cases, such as Connections
and chain-based elements. The general idea is that it's a lot more
efficient to make the decisions once in create_plan than to keep doing
this huge if/else chain in the wrapper. Just choose the right wrapper up
front. It'll be most apparent performance-wise in the case of whichever
element context is switched to first for each iteration, since the whole
wrapper setup is done for every iteration.
The tricky part is that there is now a bit of overloading of the function
pointers in a pad. The current meanings (possibly to change a bit more
soon) are:
chainfunc: as always, chainfunc pointer is mirrored between peer pads
(this may change, and the chain func may end up in pushfunc)
pushfunc: SrcPad: gst_pad_pushfunc_proxy, cothread_switch to peer
SinkPad: none (may take over chainfunc, see below) pullfunc:
SrcPad: Src or Connection's function to construct buffers
SinkPad: gst_pad_pullfunc_proxy, cothread_switch to peer
There are a number of issues remaining with the scheduling, not the least
of which is the fact that Connections are still dealt with the old way,
with _push() functions and such. I'm trying to figure out a way to unify
the system so it makes sense. Following the scheduling system is hard
enough, trying to change it is murder.
Another useful scheduling addition, mentioned above, is COTHREAD_STOPPING.
It's an element flag that's used to signal whatever code is running in
cothread context that it should be finishing up and exiting soon. An
example of this is in plugins/cobin/spindentity.c. All the loops should
now be composed of do/while loops, rather than while(1) loops:
do {
buf = gst_pad_pull(spindentity->sinkpad);
gst_pad_push(spindentity->srcpad,buf);
} while (!GST_ELEMENT_IS_COTHREAD_STOPPING(element));
The reason for this is that COTHREAD_STOPPING may be set before the above
loop ever gets started. It wouldn't do for the body of the loop to never
once get called, that would simply stall the pipeline. Note that only the
core library code is ever responsible for setting and unsetting this flag.
All elements have to do is respond to it by cleanly exiting the loop and
the function holding it.
This is needed primarily to allow iterations to occur properly.
Basically, there's a single entry point in the cothread scheduling loop,
gst_bin_iterate_func() simply switches to this cothread. If the element
in this context is allowed to loop infinitely, nothing would even switch
back to the context from which the iterate() was originally called. This
is a bit of a problem. The solution is for there to be an implicit switch
back to the originating context. Now, even I'm not sure exactly how this
works, but if the cothread that's switched to actually returns, execution
returns back to the calling context, i.e. iterate_func().
COTHREAD_STOPPING is therefore set just before switching into this
(currently randomly chosen) context, on the assumption that it will return
promptly after finishing its duties. The burden of clearing the flag
falls to the various wrapper functions provided by the Bin code, thus
element writers don't have to worry about doing that at all (and simply
shouldn't).
Related changes:
All the sources in elements/ have been changed to reflect the new system.
FIXMEs:
1) gstpipeline.c calls gst_src_push at some point, dunno why, it's
commented out now.
2) any other sources, including vcdsrc, dvdsrc, and v4lsrc will break
badly and need to be modified to work as pull-based sources.
Original commit message from CVS:
Changes made to the DEBUG system. New header file gstdebug.h holds the
stuff to keep it out of gst.h's hair. DEBUG prints out the process id,
cothread id, source filename and line number. Two new macros DEBUG_ENTER
and DEBUG_LEAVE are used to show the entry and exit of a given function.
This eventually might be used to construct call trace graphs, even taking
cothreads into account. This would be quite useful in visualizing the
scheduling mechanism.
Minor changes to various debug messages.
Also sitting in gstdebug.h is a prototypical DEBUG_ENTER that's capable of
performing DEBUG_LEAVE automatically. It does this by utilizing a
little-known GCC extension that allows one to call a function with the
same parameters as the current function. The macro uses this to basically
call itself. A boolean is used to ensure that when it calls itself it
actually runs the body of the function. In the meantime it prints stuff
out before and after the real function, as well as constructing a
debugging string. This can be used eventually to provide call-wide data
on the DEBUG lines, instead of having to replicate data on each call to
DEBUG. More research is needed into how this would most cleanly be fit
into some other chunk of code, like GStreamer (I think of this DEBUG trick
as a separate project, sorta).
Unfortunately, the aforementioned DEBUG trick interacts quite poorly with
cothreads. Almost any time it's used in a function that has anything
remotely to do with a cothread context (as in, it runs in one), a segfault
results from the __builtin_apply call, which is the heart of the whole
thing. If someone who really knows assembly could analyze the resulting
code to see what's really going on, we might find a way to fix either the
macro or the cothreads (I'm thinking that there's something we missed in
constructing the cothreads themselves) so this works in all cases.
In the meantime, please insert both DEBUG_ENTER and DEBUG_LEAVE in your
functions. Be sure to put DEBUG_ENTER after your variable declarations
and before any functional code, not to put the function name in any DEBUG
strings (it's already there, trust me), and put a DEBUG_LEAVE if you care
enough.
Changes are going to happen in the way DEBUGs and other printouts occur,
so stay tuned.
Original commit message from CVS:
Added a check to see if the plugin actually exists before even bothering
to try to load it. Saves a fair amount of debugging spew with
_gst_plugin_spew enabled.
Original commit message from CVS:
Added multiple types to the pads.
Added first preview of the capabilities system.
Autoplugging is seriously broken with these (and other) changes.
Original commit message from CVS:
Enhanced debugging by making DEBUG() print out the cothread ID as well as
the process ID. cothread_getcurrent returns the current cothread ID, or
-1 if cothreads aren't set up in this process context.
Original commit message from CVS:
Rearranged cothread_switch a bit to move all the code for error conditions
to the end, using goto's to get there. This has the presumed advantage of
consolidating all the normally run code into one chunk, reducing jumps
(and the associated penalties in any modern processor) and limiting cache-
line usage. It may be instructive to look at the generated assembly for
this revision and the previous, to see if gcc is smart enough to do this
for us anyway.
If you want to turn off some of the checks (they are all checks for NULL
pointers, if you're curious) for a speed gain, disable the #define of
COTHREAD_PARANOID at the top.
Original commit message from CVS:
Added code to force the gsttypes plugin to load before gstelements, by
simply having gstelements.c request the plugin. Solves some dependency
problems. This is the correct method of doing this for now, though I had
a thought:
Have a static list of hard dependencies that the plugin system is responsible
for satisfying before even trying to load the plugin. Makes plugin design
easier.
