April 2006 - MPlayer-dev-eng - lists.mplayerhq.hu

v4l/v4l2 buffer size bug
by Alan Curry 25 Apr '06

25 Apr '06

This code in tvi_v4l.c and tvi_v4l2.c: sysinfo(&si); if (si.totalram<2*1024*1024) { bufsize = 1024*1024; } else { bufsize = si.totalram/2; } doesn't work correctly when si.mem_unit!=1 which will be the case whenever any element of the sysinfo is bigger than 4G. The fix would be to substitute si.totalram*si.mem_unit for si.totalram. mem_unit hasn't always been there. Detecting whether the running kernel supports mem_unit, and doing the right thing when <sys/sysinfo.h> contains the old struct definition without mem_unit, is left as an exercise for the v4l maintainer.

2 2

Change pts values to use doubles instead of floats
by Uoti Urpala 25 Apr '06

25 Apr '06

Floats do not have sufficient precision to hold pts values in long streams/files. After 8192 seconds the distance between two consecutive floats is about 1 ms, and the error grows proportional to file length, so after 65536 seconds (less than a day) it's 8 ms. The attached patch changes the common data structures to doubles. It does not yet change any demuxers to produce pts values with higher precision, many of them use floats internally.

8 14

[PATCH] win32 gui
by Erik Lunchpail 25 Apr '06

25 Apr '06

Greetings, I was directed to post this here from the mplayer-cygwin list, so here goes. This is a heavily heavily modified version of Sascha Sommer's original win32 gui for win32. I began modifying it quite a while ago, with the help of Gianluigi Tiesi who has been kind enough to host binaries and fix my bugs. It's always been available on his site (http://oss.netfarm.it/mplayer-win32.php) which is also linked to from the mplayer homepage under related projects. It uses standard win32 api calls to emulate the current linux gui, so it works in both cygwin and mingw. The goal was to try and handle everything that we possibly could within the gui source, and to stay out of the external mplayer source tree as much as possible. It takes advantage of the gui code that is currently in place within the source tree, so only very minor patching is necessary. Only 3 files require patching, configure, Gui/Makefile, and mplayer.c. The former 2 are self explanitory, but i'll briefly explain the mplayer.c changes. Basically it works the opposite in windows as it does in linux (go figure huh?). The gui is enabled by default, however it is disabled when called via slave, so this still allows it to be used by any frontends around that call mplayer's slave functions. Secondly, i've added a COPYDATA struct which allows for IPC which is handled within gui.c. Lastly, since the sub window is fairly new, and is handled differently than the linux gui, a small 3-liner to allow the sub window to correctly adjust to new aspects when using runtime aspect switching. Some screenshots can be viewed here: http://vayne.fdns.net/mplayer/shot1.png http://vayne.fdns.net/mplayer/shot2.png http://vayne.fdns.net/mplayer/shot3.png Some features: IPC with WM_COPYDATA (as stated above) Fully skinnable with any supported mplayer skin. Playlist loading/saving/shuffling. Full drag&drop support (including playlists) DVD/vcd/cd support (the latter with libcdio) Aspect switching support, whether with -aspect on the cmdline or via runtime aspect switching. DVD title/chapter switching dialog. Preferences dialog. Skin browser. Support for loading external subtitle files. Systray support. Console toggle for stdout/stderr debug info. The sub window takes advantage of WinID when using vo directx, but will hide when using other vo's. And as well, the sub window should automatically hide when a file is audio only and only pop back up when we have video. Also included is an optional (to be applied last) patch for DVD menu's for use with Atvos Attila's dvd menu patches. Questions? comments? Regards, Erik __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

8 27

[PATCH] simplify x86 CPU flag/extension check
by Diego Biurrun 25 Apr '06

25 Apr '06

The attached patch works fine for me on my K6-III, but I'd appreciate some testing on other processors. Compare config.h and config.mak with and without the patch, also when passing options like --disable-mmx etc. Diego

3 5

[PATCH] printf -> mp_msg
by Ötvös Attila 25 Apr '06

25 Apr '06

Hi All! Files: find_sub.c, sub_cc.c, subreader.c liba52: parse.c, ao_oss.c, ao_sun.c, audio_out.c libdha: pci.c, sysdep/libdha_win32.c, sysdep/pci_win32.c libmenu: menu.c, menu_cmdlist.c, menu_console.c, menu_filesel.c, menu_param.c, menu_pt.c, menu_txt.c, vf_menu.c libmpcodecs: ad_libdv.c, mp_image.h, pullup.c, vd_dmo.c, vd_dshow.c, vd_ffmpeg.c, vd_realvid.c, vd_vfw.c, ve_lavc.c, ve_raw.c, ve_vfw.c, vf.c, vf_crop.c, vf_cropdetect.c, vf_expand.c, vf_fame.c, vf_format.c, vf_framestep.c, vf_noformat.c, vf_palette.c, vf_phase.c, vf_test.c, vf_tile.c, vf_yuy2.c, vf_yuy9.c libmpdemux: cache2.c, cddb.c, demux_asf.c, demux_audio.c, demux_demuxers.c, demux_film.c, demux_gif.c, demux_mov.c, demux_nuv.c, demux_real.c, demux_viv.c, demux_xmms.c, stream.c, tvi_v4l2.c, video.c libvo:aspect.c, mga_common.c, vesa_lvo.c, video_out.c, vo_3dfx.c, vo_dxr3.c, vo_mga.c, vo_null.c, vo_png.c, vo_sdl.c, vosub_vidix.c, vo_svga.c, vo_syncfb.c, vo_tdfxfb.c, vo_tdfx_vid.c, vo_tga.c, vo_vesa.c, vo_x11.c, vo_xv.c Best regards! Attila

8 23

HEADER_SEARCH_SIZE too small in demux_nsv.c ?
by Soeren Sonnenburg 24 Apr '06

24 Apr '06

Hi List, I realized that most of my .nsv streams just don't work... I increased the HEADER_SEARCH_SIZE and printed out the index i when NSV{s,f} was found and it often was >65k... sometimes even >200k ... so to be on the safe side I put #define HEADER_SEARCH_SIZE 1920000 in the file... however sth. like 256k might be enough, not sure what the specs say. Can anyone comment on this and fix this in cvs ? Soeren

3 7

[PATCH] Updated audio and sync handling related patches
by Uoti Urpala 24 Apr '06

24 Apr '06

1 - Remove kernel HZ=100 assumptions that are no longer true from timing. 2 - Fix demux_mkv.c to not give bogus timestamps for laced packets with no default duration. 3 - Move a definition of MP_NOPTS_VALUE from muxer.h to demuxer.h to make it available in more files (needed for next patch). 4 - Implement a better way to calculate current audio pts and use it for audio decoded with ad_libvorbis, ad_ffmpeg and ad_faad. 5 - Drop more buffered audio data left from old position when seeking. 1-4 are slightly changed from previously posted patches or updated to match current CVS (no significant functionality changes), 5 is new.

5 4

FFmpeg include header
by Ivan Kalvachev 24 Apr '06

24 Apr '06

Here is an early patch, it uses trick similar to what FreeType does to declare include header paths. You may have noticed that there are quite many places where #ifdef are used to include ffmpeg headers from default system location or from internal mplayer one. This code duplication could be removed if we #define the path string and include the define. The patch I created remove about 27 lines total, please take a look and test it. At least I was able to compile mplayer with this patch applied. I would like to hear your recommendations: the names of the defines. FF_AVCODEC_H is too similar to the AVCODEC_H used by avcodec.h itself. I was thinking about FF_AVCODEC_HEADER (too long but..). Do you have better ideas? jpeg_enc.c and vf_zrmjpeg.c are unchanged by my patch. They include quite many internal headers and i think they behave more like parts of lavc. I'm not going to mess with them ;) vf_qp - does it really need all these heavy includes. I think we can remove at least the dsputil one. After this patch is committed I will make another for avformat and postprocess. Let's talk about code.

3 6

[PATCH] GUI: fix audio/video id in gtk/menu.c
by Stanislav Maslovski 24 Apr '06

24 Apr '06

Hello, When playing an ogg file in gmplayer I noticed that changing audio streams from the menu was buggy. An investigation showed that in gtk/menu.c the audio and video streams' IDs were assumed to be equal to the indexes of the corresponding demuxer->a_streams[i] and demuxer->v_streams[i] members. However, from reading demux_ogg.c and demux_mkv.c codes I see that those IDs are simply continuously counted from 0. Please, correct me if I am wrong or if there are other cases I have not noticed. The patch is attached. -- Stanislav

5 8

[PATCH] if(verbose>X) --> mp_msg_test
by Diego Biurrun 24 Apr '06

24 Apr '06

Hi, I was staring at progress bars a lot at work yesterday, so I did some janitorial work that should help the pending mp_msg patch from attila. I've converted all the 'if (verbose > X)' calls to mp_msg_test, which is the proper way and gets us rid of a few global vars as an added bonus. I intend to apply this tomorrow, shout now if I messed up somewhere. Diego

4 22