The number of crash reports on Linux make it obvious that 0.8.5 and 0.8.4 are not binary-compatibile, so we should definitely change the entry point name
* updated and partially (re-)added the Georgian (by Alexander Didebulidze), Hebrew, Hungarian, Hindi, Russian, and Italian (all by Rosetta translators) localisations
* Fix subtitle alignment if we don't have styles. fixes #619
I chose to re-add the align paramter to the subpicture region struct, since adding a style struct for every single subpicture for just an alignment seemed to be overkill to me.
I was actually trying to understand why command line option parsing on my computer with self compiled VLC behaves as "If the first character of optstring is '+' or the environment variable POSIXLY_CORRECT is set, then option processing stops as soon as a non-option argument is encountered." while neither of these conditions is true (see the while( getopt_long( ... ) != EOF ) loop).
Small illustration:
./vlc -vvv -I dummy input.ts --color
All options before the first non option argument work as expected (-vvv and -I dummy). All options after the first non option argument are also interpreted as being non option arguments.
Help would be appreciated.
(I've already had reports of other people having the same kind of issues)
* Fix ffmpeg libavformat linkage when we don't use --with-ffmpeg-tree (this fixes Flash playback)
* Add Shorten codec. This refs #632 It doesn't seem to work, but that might be an endianness issue. will someone please check with this sample? http://www.mplayerhq.hu/MPlayer/samples/A-codecs/lossless/luckynight.shn
Win32 specific changes: Add --started-from-file option to tell VLC that it was started by clicking on a file (or disc). Add --one-instance-when-started-from-file option to tell VLC to use one instance mode when --started-from-file is used. See replies to r15125 for an explanation.
* store the value of 'macosx-black' as one of VLCWindow's properties to prevent rare crashes when en-/disabling vout-filters or leaving/entering fullscreen-mode really fast.
When going to fast to fullscreen/windowed-mode VLC can still crash (now at vout.m:375), but that scenario isn't realistic because you need to perform this function at least 5 to 10 times per second to get this done.
* Use run-time detection of UTF-8 as current charset instead of hard-coding to be if and only if Mac OS X
* Use CP1252 as a fallback instead of ISO-8859-1 (as far as printable characters are concerned, the first is a superset of the second, and it is the most commonly used charset in the west to date)
all: for Cygwin, make sure that the program-prefix is empty, so that libgcrypt-config & gnutls-config are not prefixed by the target build (i586-mingw32msvc), which causes prevents configures scripts from detecting them
extras/contrib/src/Makefile: cleaned up INSTALL_NAME macro for MacOS X: all the dylib in contrib/lib are soft-linked to their distributed version in 'vlc-lib'. this way, VLC build system can use contrib libs on MacOS X the same way as for UNIX, and should simplify porting to libtool
configure.ac: cleaned up Cocoa.h detection, removed link path dependency to 'vlc-lib' as it is no longer necessary
Makefile.am: fixed installation of dylibs as vlc_ prefix is already appended by INSTALL_NAME
NOTE: this change is highly disruptive for MacOS X build, I strongly recommend that you rebuild the contrib packages for this platform. Other platforms are unaffected
M /trunk/modules/gui/wxwidgets/dialogs/playlist.cpp
We can now DnD a file everywhere on the playlist dialog. Droping above the tree will insert at the begining of the playlist. Droping under the tree will insert at the end of the playlist.
* MainMenu.nib, playlist.m: use NSSearchField instead of NSTextField, so we don't need to replace it manually anymore. This breaks the 10.2-compatibility. (refs #189)
* update.m: use mtime() instead of NSCalenderDate for the time-stamp
* don't store the temp release note in the user's Caches-folder, but in /tmp. Renamed it to vlc_releasenote_%@.tmp, whereas %@ is a timestamp (%m-%d-%y--%I.%M.%S.%F)
* download the updated versions through VLC and not through your default browser
* removed a crasher from interaction.m
Note: the progress-window used to show the progress is quite hacky at the moment and will be replaced through a good implementation once the interaction-core is finished.
M /trunk/extras/MacOSX/Resources/English.lproj/MainMenu.nib/classes.nib
M /trunk/extras/MacOSX/Resources/English.lproj/MainMenu.nib/info.nib
M /trunk/extras/MacOSX/Resources/English.lproj/MainMenu.nib/keyedobjects.nib
M /trunk/modules/gui/macosx/controls.h
M /trunk/modules/gui/macosx/controls.m
* added a "Go To Specific Position" feature (closes #377)
- accessible by double-clicking on the time-counters and through a menu-item in "Playback"
- the class of the time-counters is VLCTimeField now, because we need to fetch the mouseDown-event. There are no further changes compared to NSTextField.
all: renamed DLL import library as libvlc.dll.a, as this is what mingw/cygwin ld expects to see. for more info refer to: http://www.sourceware.org/binutils/docs-2.16/ld/WIN32.html
all: for WIN32, create an import library libvlc-dll.a when linking libvlc.dll; updating linking for vlc.exe and activex.dll to use import library in place of dll.
Fix running vlc from path in windows (see http://forum.videolan.org/viewtopic.php?t=17435 ). This could break the mozilla plugin on windows (see http://trac.videolan.org/vlc/changeset/9106) so i'm commiting and i'll do a nightly to check. One way to be sure would be to ask Gibalou but ....
M /trunk/extras/MacOSX/Resources/English.lproj/Wizard.nib/classes.nib
M /trunk/extras/MacOSX/Resources/English.lproj/Wizard.nib/keyedobjects.nib
M /trunk/modules/gui/macosx/wizard.h
M /trunk/modules/gui/macosx/wizard.m
* support for applying the same settings to multiple files in single run of the wizard (just select multiple items in the playlist)
- note: due to this, the types of some objects in the main dictionary changed from NSString to NSArray. Additionally, this code does neither run nor compile on 10.2 anymore.
- various loops are being performed during the run of the wizard now. Please shout if this is noticeably slower on older Macs
* solved a UTF-8-issue which prevented the usage of certain files which were added to the playlist previously
* optimised the existing code here and there
* make sure that all settings are kept for a session
* Improved renderering of YUVA/P onto YUV variants. refs #539
I420 still not fixed for this bug.
A simple average of the UV pixels, wasn't giving satisfying results, so I added a bit more logic, to decide based on the transparency value if UY averaging is desireable for the pixels. Produces nice results now, with a very slight shadow somtimes to the right side of the images/glyphs. Looks quite good.
* Fix the duration of the dvdread module. refs #198.
Now duration is accurate, but current time is still block based, so on VBR discs the duration of a second may have some weird variations. Getting accurate time is possible (see what was reverted in [12862]), but this will break the relative SET_TIME (might actually go seeking in the wrong direction for instance). The correct fix here is to expand dvdread with a method that checks the TMAPTI IFO table of a Title, to find what block approximately goes with a certain time. (the TMAPTI table is usually in .5 secs interpolation might possibly be required as well) As a backup an estimate can be made using the ADMAP table.
The behaviour of the dvdread module is now equal to that of the dvdnav module when it comes to time and seeking.
* Don't put huge pictures on the stack. It is limited on some platforms. use malloc.
* Use the correct pitch values when reading from and writing to pictures (out pitch can differ from in pitch, and there is also visible pitch and visible lines.
M /trunk/extras/MacOSX/Resources/English.lproj/Wizard.nib/classes.nib
M /trunk/extras/MacOSX/Resources/English.lproj/Wizard.nib/info.nib
M /trunk/extras/MacOSX/Resources/English.lproj/Wizard.nib/keyedobjects.nib
M /trunk/modules/gui/macosx/wizard.h
M /trunk/modules/gui/macosx/wizard.m
* clean up (mostly replacements of nestings of stringByAppendingString by stringWithFormat)
* removed the stupid "wizard help window" and replaced the help sheets with calls of NSBeginInformationalAlertSheet. This might be beautified with NSAlerts in the future, since we don't care for 10.2 anymore.
* basic support for SSA formatted subs. (disable with --no-subsdec-formatted option)
* changes to freetype renderer to support the font color, size and alpha options of text_style_t (other options are possible, just not implemented yet. Full text_style_t support requires a more advanced renderer though. )
* changes to modules to support text_style_t instead of the old sub options.
* Some changes to subsdec to only iconv to UTF-8 if source is !UTF-8.
refs #82
not supported are style runs (styles on substrings).
* Disable the non functional QT/Quartz output on the intel machines. We should first fix it on OSX, and currently it's annoying the intel debugging proces.
* Open the .sub file of the vobsubs, with the ToLocale version of the path. Should fix issues where the path of the vobsubs contains non-ascii chars. The .sub could then not be opened in some cases since it would try to open with the UTF-8 path. (We should check VLC modules for other cases of this annoying problem)
misc/vlm.c : export some p_input variables (state, position, time, length, rate, title, chapter and seekable) when using the vlm "show (media)" command.
r13909 | jpsaman | 2006-01-14 10:36:16 +0100 (Sat, 14 Jan 2006) | 1 line
Changed paths:
M /trunk/modules/access/dshow/common.h
M /trunk/modules/access/dshow/dshow.cpp
Revert revision 13903. It is implemented in a different way by checking if the option --dshow-chroma is set. If it is set then the chroma is forced, otherwise it is not. This should solve the regression of previous commit, by letting users specify the chroma type to use. To get the previous default behaviour specify IV420 as preferred chroma type either on the commandline or in the Capture Device advanced tab.
Patch by Paul Sokolovsky <pmsicml _AT_ gmail _dot_ com>.
Remove unconditionally enforcing of "preferred" media type upon the capture devices. Some devices don't handle this properly, which often results in a BSOD. By not setting a preferred media type the problem dissappears. Please check this with other type of devices for any regression.
