useful sources of data. This *may* help solve the issue of what test data to
use.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4047
Use of ESOffset sorted out a bit more -- I think it now makes sense.
And since I can't find any way for doctest to "see" C stdout/stderr,
just make rundoctest.py output an explanatory message (and the RED/GREEN
thing just makes it a bit quicker to ignore worrying texts).
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4046
read *or* write. Thus change the way it is actually opened for read.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4037
Default value is 25.00, so if no value is given the code computes everything exacly as before.
The value of the frame rate is used to compute the gop duration.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4033
esmerge
2. Do not output spurious packets if in tsdirect mode in tsserve.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4030
using the (correct) close_elementary_stream. This is tolerant of a NULL
argument (because it's designed to pair with open_elementary_stream,
which can fail).
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4026
Python package will have the same name as the pyx file (or nasty things
happen when you try to import it), and I ultimately want the Python module
to be called "tstools".
Add a (probably) temporary Makefile for pyrex, and the Setup.py that does
the actual work.
On my Mac, this how builds something I can import into Python, and then
discover errors in... so fix the 'filename' attribute of ESStream, both
to exist (!) and secondly to be visible to Python.
--HG--
rename : es.pyx => tstools.pyx
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4025
file which has caused problems. No, that's not a proper solution.
Added some XXX comments discussing that.
Also, added some code to allow for reading beyond the last PCR, and using
the last-determined rate. This isn't really tested yet.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4021
stop inappropriate writing of (extra, possibly wrong) program data
when we are mirroring TS packets directly.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4018
be being used for such (otherwise, it can be very unobvious that this
is so).
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4016
mode distributed with Vim -- I actualy prefer rest.vim, but that's an
extra one has to download onself).
The appropriate [X]Emacs line is left for the moment, since I don't
currently use [X]Emacs for editing these files.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4014
attempt to autodetect whether it is running on Mac OS X). Thus it replaces
the original Linux Makefile (and we're back to a GNU make file plus an
nmake make file for Windows).
--HG--
rename : Makefile.osx => Makefile
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4011
make
make clean
did not delete all the binaries. Now, target objclean does what clean
used to do, and clean/distclean act more conventionally.
Also, Makefile.osx *should* now be a suitable makefile for Linux/BSD
as well -- but wait until I can test it at work (on Linux) before
making it the "main" Makefile.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%4010
In choose_nal_type in esdotc.c there is a boolean value rec_point_required. It you set it to FALSE any I pic is considered a random access point.
The field recovery_frame_cnt is not used (as if it was 0). If it is != 0 a warning message is printed.
The fields broken_link_flag, exact_match_flag and changing_slice_group_idc are not considered (as 0).
In the final message the max, min and mean duration of a GOP are shown. For H.264 streams this is defined as the distance between 2 random access points. If no full GOP is found this info is not displayed.
The -gop switch works as before.
In esdots a function choose_nal_type has been added.
In nalunit.c a function read_SEI_recovery_point has been added.
The structure nal_SEI_recovery_data has been added, in order to cope with the recovery_point message. It's added to nal_innards too. See nalunit_defns.h
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%409
- For MPEG-2 streams it outputs informations about the size of a Gropu Of Pictures (GOP): min, max and average duration in sec. Video is supposed to be @25fps.
- For AVC streams, where the GOP is not formally defined, it computes the same date looking at the interval between two I or IDR frames (whichever comes first).
Using the switch -gop display the duration of each GOP (or pseudo-GOP fir AVC)
NOTE that a new function hab been added:
static char choose_nal_type(access_unit_p access_unit, int verbose);
Called by dots_by_access_unit, it returns a single character which specifies the type of the access unit.
+ tools.txt has been modified describing this new functionality and switch.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%408
(with -v) didn't match the code, and the whole thing was rather wrong.
--HG--
extra : convert_revision : svn%3Aeff31bef-be4a-0410-a8fe-e47997df2690/trunk%404
tibs