Speex may decide not to consume any samples because it can't write any. I've
seen a hang during draining caused by the resample loop never terminating.
In that case, resampling happened as normal until olen was 0 but ilen was
still 1. _process_native then reduced ichunk to 0, so ilen never decreased
below 1 and the loop never terminated.
Instead of reverting 684cf44 ({audioresample: don't skip input samples),
break only if all output samples have been produced and speex refuses
to consume any more input samples.
https://bugzilla.gnome.org/show_bug.cgi?id=732908
when downsampling, the output buffer can be filled before all the input
samples are consumed. this is correct: when downsampling, several input
samples are needed for each output sample, so when only a small number of
input samples are available the number of output samples produced can be 0.
the resampler, however, was discarding those extra input samples instead of
clocking them into its filter history for the next iteration. this patch
fixes this by removing the check that the output buffer is full. the code
now always loops until all input samples are consumed, and relies on the
calling code to have provided a suitably sized location for the output.
note that there are already other checks in place in the calling code to
ensure that this is the case.
https://bugzilla.gnome.org/show_bug.cgi?id=732908
This is an adaptation of patch #3 from Jyri Sarha
( http://lists.xiph.org/pipermail/speex-dev/2011-September/008240.html ),
but without the NEON optimizations (these come in a separate commit).
The idea is to replace SATURATE32(PSHR32(x, shift), a) operations with a
combined SATURATE32PSHR(x, shift, a) macro that can be optimized for
specific platforms (and also avoids rare rounding errors).
Signed-off-by: Carlos Rafael Giani <dv@pseudoterminal.org>
Original commit message from CVS:
* gst/audioresample/debug.h:
* gst/audioresample/resample.c: (resample_init):
Since I really am not interested in a debug line for each sample
being processed, move the library's debugging to its own category,
libaudioresample
Original commit message from CVS:
2006-10-10 Zaheer Abbas Merali <zaheerabbas at merali dot org>
Patch by: Josep Torre Valles <josep@fluendo.com>
* ext/gnomevfs/gstgnomevfssink.c:
* ext/gnomevfs/gstgnomevfssrc.c:
Fix URI interface implementation return type.
* ext/pango/gsttextoverlay.c: (gst_text_overlay_set_property):
Fix what looks like a copy/paste issue when assigning values.
* gst-libs/gst/audio/gstaudiofiltertemplate.c:
(gst_audio_filter_template_get_type):
Cast to prevent Forte warnings.
* gst-libs/gst/cdda/gstcddabasesrc.c: (gst_cdda_base_src_create):
Fix URI interface implementation return type.
gst_pad_query_position requires a signed integer pointer as
3rd parameter, GstClockTime is unsigned.
* gst/audioconvert/audioconvert.c:
Fix integer overflow when treated as signed.
* gst/audioresample/resample.c: (resample_add_input_data):
Cast to prevent warnings on Forte.
* gst/ffmpegcolorspace/imgconvert.c: (build_rgb_palette):
Fix integer overflow when treated as signed.
* gst/ffmpegcolorspace/imgconvert_template.h:
Fix integer overflow when treated as signed. RGBA_OUT shifts bits.
* gst/playback/gstdecodebin.c: (queue_filled_cb),
(cleanup_decodebin):
Who initialises a guint to -1!
Cast function pointers to prevent warnings on Forte.
* gst/playback/gstplaybasebin.c: (queue_deadlock_check),
(queue_threshold_reached):
Cast function pointers correctly to prevent warnings on Forte.
* gst/playback/gststreaminfo.c: (gst_stream_info_dispose):
Cast function pointers correctly to prevent warnings on Forte.
* gst/subparse/gstssaparse.c: (gst_ssa_parse_setcaps):
Obvious change to unsigned, 0xEF > max signed char.
* gst/tcp/gstmultifdsink.c: (get_buffers_max), (count_burst_unit):
GstClockTime is unsigned, initialise correctly.
* gst/tcp/gsttcp.c: (gst_tcp_socket_write):
Cast so pointer arithemetic doesn't cause warnings on Forte.
* gst/videorate/gstvideorate.c:
Use correct return value.
* tests/examples/seek/scrubby.c:
GstClockTime is unsigned, initialise correctly.
Original commit message from CVS:
* gst/audioresample/Makefile.am:
* gst/audioresample/debug.h:
* gst/audioresample/gstaudioresample.c:
* gst/audioresample/resample.c: Convert to using gst debugging
Original commit message from CVS:
* gst/audioresample/debug.c:
* gst/audioresample/gstaudioresample.c:
add room for extra overlap samples when asked to transform size
protect against possible mem corruption and check for discrepancies
between written size and outbuffer's size so we can warn for
potential problems
* gst/audioresample/resample.c: (resample_init),
(resample_get_output_size_for_input), (resample_get_output_size),
(resample_set_n_channels), (resample_set_format):
set debug level based on RESAMPLE_DEBUG env var
make sure that get_output_size* returns a whole number of
sample_size
set sample_size each time either channel or format is set
* gst/audioresample/resample_chunk.c: (resample_scale_chunk):
* gst/audioresample/resample_functable.c:
(resample_scale_functable):
* gst/audioresample/resample_ref.c: (resample_scale_ref):
remove r->sample_size, it's done in resample.c now
add some debugging to the ref implementation
make sure we only give back bytes that are wholes of the sample
size
