Tap-to-time provides the user the ability to tap to the lyrics/syllables
of the song in order to time lines or karaoke. It consists of these
extra UI interactions:
- **Indicator**: tap marker: a designated marker that can be moved to
the current audio position; indicated in:
- the audio display by a green arrow underneath a marker
- the karaoke display by a green-colored syllable
- **Control**: tap marker: the tap marker can be changed by selecting
syllables on audio display in karaoke mode, or clicking the markers on
audio display in dialogue mode
- **Control**: ctrl-right-click audio display: starts playing the audio
from that exact position until the end of the file
- **Option**: Timing/Tap To Time: enables the tap marker indicator and
commands
- **Button**: time_opt_tap_to_time: toggles the Timing/Tap To Time option
- **Button**: time_tap_connect (hotkey I): a command that:
- moves the tap marker's position to the current playing audio
position
- sets the next marker to be the tap marker
- if the tap marker is already the last marker AND BOTH autocommit AND
next-line-on-commit is ON, will move onto the next line
- if moved on to the next line, also sets the start marker to the current
audio position, so the two lines are connected, and moves to the
next tap marker (essentially reinvoking time_tap_connect once)
- **Button**: time_tap_no_connect (hotkey O): similar to
time_tap_connect, except it will not set the next line's start
position even if moved to the next line
Expected workflow:
1) User loads song lyrics
2) User splits each line into syllables
3) User turns on tap-to-time, autocommit, and next-line-on-commit
4) User plays audio from beginning, tapping time_tap_connect to each
syllable, occasionally tapping time_tap_no_connect when a break between
lines is desired
5) If user messes up a line, they can set the tap marker to where they
want to restart from, and ctrl-right-click to start the audio a few
seconds before it
6) Syllables can be split/merged at will, and adjustments to timing can
be done using normal karaoke timing controls
Having the timing correction in MoveMarker is extremely surprising
behavior if any other part of the code uses MoveMarker expecting it to
work correctly as advertised
Some matroska files have audio start at timestamp 0 and video later.
In this case mkvtoolnix seems to use the first block of the first
cluster to the audio track (I would assume this is only an
implementation detail and not really from the matroska specs. And also
could happen in other cases without the video being delayed, but that's
not the point). Aegisub used to read this first block and use its
timestamp as the starting point of the video track.
With this commit, Aegisub tries to read all the blocks until it can read
the first timestamp of the video track and use it for the subtitles'
timestamps. Audio tracks don't seem to be impacted by these changes.
macos-latest now runs on Apple Silicon, which fails one test
and probably still needs extra work for the packaging to work.
Until that's figured out, run on macos-13.
I was originally planning to eventually rebase the various feature
branches on top of TypesettingTools/Aegisub's master, which switched
to C++20 and added various bigger refactors, but that turned out to
be even more involved than expected (due to the various
interdependencies between branches and the additional commits on feature
complicating rebases). ICU now requires C++17 and BestSource was already
pinned to an old version since newer versions need C++17, so let's just
bump the standard and put off the actual rebasing for later...
Equivalent functionality (allowing dependencies to be meson subprojects
and fixing the zlib HAVE_UNISTD_H issue) was merged upstream now
(thanks kasper), so this is no longer necessary. dav1d is now
enabled using -Dffmpeg:libdav1d=enabled.
Update API usage, incorporate new features, and refactor a bit:
- Use new progress callback to show progress bar
- Use new frame info functions to get timecodes/keyframes instead of
manually decoding
- Use new tracklist API to show track selection dialogs
- Expose option of whether to apply RFF
- Show a warning when bestsource had to revert to linear decoding
- Reuse the swscale context across frames and ensure the format
stays constant
- Add xxhash wrap
- No longer mark the bestsource video provider as slow
[src\meson.build] Add DirectWrite has dependency
[src\font_file_lister_gdi] Rework GDI FontCollector to use DirectWrite
This replaces all the logic of using the Windows registry to obtain the font path by using DirectWrite. The goal is simply to improve the quality of the code. This doesn't change any functionality
[src\meson.build] Remove Uniscribe has dependency
Uniscribe was only used for the FontCollector. Since we now use DirectWrite, we don't need it anymore.
[src\dialog_fonts_collector] Catch exceptions that FontCollector may raise
On Windows, the initialization of the FontCollector can raise an exception
[src\font_file_lister] Document the exception that GdiFontFileLister can throw
[src\font_file_lister_gdi] Correct possible memory leak when an error occur
Fix error caused by AddFontResource on Windows 10 or higher
[meson.build] Replace add_project_arguments with conf.set for HAVE_DWRITE_3
[src\dialog_fonts_collector] Update message error and optimisation
[src\font_file_lister_gdi] Correct documentation typo
[src\font_file_lister_gdi] Cosmetic nit - Initialize hfont in one line
[src\font_file_lister_gdi] Cosmetic nit - Remove if statements brace
[src\font_file_lister_gdi] Replace WCHAR param of normalizeFilePathCase to std::wstring
[src\font_file_lister_gdi] Replace WCHAR by std::wstring
[src\font_file_lister_gdi] Use IDWriteFontFace::GetSimulations to detect fake_italic/fake_bold
See this comment: https://github.com/arch1t3cht/Aegisub/pull/107#issuecomment-1975229652
[src\font_file_lister_gdi] If Win7/8 has Win 10 SDK on compile time, correctly verify if font has character(s)
With the Visual Studio 2019 toolchain on Windows 7, it installs the Windows 10 SDK by default. Because of this, ``HAVE_DWRITE_3`` is true, so the ``QueryInterface`` always fails. Now, if the ``QueryInterface`` fails, we try to verify if the font has characters with a Windows Vista SP2 compatible code.
[src\font_file_lister_gdi] Support facename that contains only whitespace AND truncated facename
Problem 1:
Previously, if a user wrote "\fn ", it would return the font Arial, which is not what we want. This is because when we request EnumFontFamiliesEx with whitespace or an empty lfFaceName, it will enumerate all the installed fonts.
Solution 1:
To resolve this issue, let's implement a solution similar to libass to determine if the selected facename exists: 649a7c2e1f/libass/ass_directwrite.c (L737-L747)
Problem 2:
GDI truncates font names to 31 characters. See: https://github.com/libass/libass/issues/459
However, since I changed the method to determine if a facename exists, I ensured that we still support this "feature".
To test this, I used the font in: https://github.com/libass/libass/issues/710
[src\font_file_lister_gdi] Add a FIXME comment regarding the utilization of std::wstring over WCHAR
[src\font_file_lister_gdi] Add FIXME comment about charset
This can be useful when scripts want to read the entire frame
and pass it on somewhere else (say, write it to a file or use some
library function on it).
This still breaks when the 0 digit is narrower than some other
digit, but that's unlikely and the current behavior now matches
the original behavior when all times are < 10h.
Fixesarch1t3cht/Aegisub#111 .
This is what users will want in the majority of cases, and switching
to this makes using this function much easier.
However, this does break backwards compatibility. Luckily to my
knowledge this function is not actually used in any existing published
script (all scripts using get_frame just use getPixelFormatted instead)
so the damage shouldn't be too large. But this is also why I'd rather
rip off the band-aid of breaking backwards compatibility now than later.
When destroying a frame handle, the previous logic would copy the
shared_ptr in the userdata and then free it twice (once explicitly
and once at the end of the function), which is actually UB, even
if if worked fine so far. This commit now ensures that it's the
actual userdata's shared_ptr that's freed in the gc function.
An unfortunate consequence of this is that the grid widths have to
be computed on every change to line timing, which will probably
have an impact on timing performance when autocommit is enabled.
It's probably not a huge impact but it's worth keeping an eye on.
All current renderers and muxers have no issues with longer timestamps.
However, classic VSFilter keeping track of time with a 32-bit int,
leads to a more meaningful limit at 596h, so enforce that instead.
Since the stringification code for SRT and ASS (but not SMPTE) relied on
hours being single digit, replace them with a simple format string.
GUI inputs are still limited to <10h, but at least
pre-existing timestamps no longer get mangled.
The vfr class (understandably) does not like timecode lists with only
one element, so the current default script breaks when opening images.
So to not bloat the default script too much we just ignore
__aegi_timecodes in the provider in these cases.
This reverts commit c487dd2bdb.
Revert the commit that
partially reverted fffb138b81 ...
It turns out that this very much did break IME input, so reverting
this for now. This does make Aegisub use private symbols in wx again,
but the CI builds build wx from the wrap now. Still, this is why
this is on the workarounds branch. If this breaks for too many
people this could either go behind a meson option or be fixed in
some better magical way that I have yet to find.
Fixesarch1t3cht/Aegisub#90Fixesarch1t3cht/Aegisub#91
This happens via VapourSynth's message logger.
Also add wrapper functions for this to aegisub_vs.py and add progress
updates everywhere - both to improve UX and to help with debugging for
when scripts get stuck somewhere.
On Windows, plugins are shipped with Aegisub now and loaded from
aegisub_vs.py, so user plugins installed out of Aegisub are more
likely to just interfere with that (like when the user has an
old version of libvslsmashsource installed).
So this option defaults to off on Windows, but to on everywhere else
since VS plugins aren't shipped with Aegisub there.
Note that this option only disables autoloading of user plugins.
VapourSynth loads system plugins no matter what.
- Fix selection not aborting when the preferred provider returns null
- Fix the default video provider being "ffmpegsource" (lowercase)
instead of "FFmpegSource", which would trip up provider selection
- More generally, make sure the preferred video provider actually exists
and fall back to the default (FFmpegSource) if not.
Fixesarch1t3cht/Aegisub#23 .
Fixesarch1t3cht/Aegisub#61 .
Fixesarch1t3cht/Aegisub#83 .
Ubuntu doesn't have ffmpeg 6.0 yet, but we want to bundle ffms2 master
since the distro version still has seeking issues. Compiling the ffmpeg
wrap needs nasm.
Also bump libdav1d and turn the wrap patches into diffs. This requires
bumping the minimum meson version to 0.63, but meson 1.0 is available
pretty much everywhere now so this should be fine.
IsConversionSupported unconditionally calls iconv_close on the
descriptor returned by iconv_open. This may result in crashes if
iconv_open returns iconv_invalid.
This became necessary now that more providers were added. Providers can
be proritized for certain file types (e.g. .vpy files will always be
opened with VapourSynth), and when the default provider fails on a file,
the user will be notified and be asked to pick an alternative provider.
The CI builds link against XAudio2_9, which isn't available on Windows 8
and lower. As a quick and dirty workaround, this commit ships the
XAudio2 redistributable on those older Windows versions. Simply renaming
the redistributable to XAudio2_9.dll isn't strictly allowed by the
documentation, but it works and is probably an acceptable hack to
support an outdated operating system...
Fixesarch1t3cht/Aegisub#16 .
Replacing all uses of LuaToAssEntry with LuaToTrackedAssEntry
also replaced its use in LuaParseKaraokeData, which would cause a double
free when canceling a script after calling parse_karaoke_data.
Instead, count how many consecutive times the entry has been found to be
unused and delete it once that count exceeds a limit. This will prevent
excessive reallocating of extradata ID's in applications like folding.
This wasn't necessary before since the internal representation of
folds is be checked for consistency after each commit, but after the
switch to extradata fold operations would leave the extradata in an
invalid state. This isn't technically a problem, but it does leave more
extradata entries lying around than necessary, and it can trip up
automation scripts that aren't prepared for inconsistent fold state.
These commands were revamped in 0ef9963 but the default hotkeys were
never updated. The hotkeys were automatically migrated, but resetting
the settings back to defaults would still set invalid settings.
This would cause an assertion failure in functions like lua_for_each
when the given closure throws an error and thus leaves some values on
the stack. This can make Aegisub crash entirely instead of just catching
and reporting the error. Instead, these stack_checks can be done
manually.
While meson already provides the aegisub-update-po target
for this exact purpose, it insists on also running meson’s
aegisub-pot target first, leading to missing translation strings.
This fixes a bunch of misspellings, borked C format strings and
increases translation coverage from 1288/1612 to 1516/1612
where still missing strings mostly don’t exists in wangqr’s build.
Original commit:
98abe53b15
Using make_pot.sh and
msgmerge --no-fuzzy-matching --sort-by-file $i.po aegisub.pot
since meson’s aegisub-pot target are still not covering all
translatable strings yet.
Base POSIX leaves the handling of backslahes in the operand
implementation defined. The XSI extension specifies several
escape sequences, like e.g. \n, which shall be transformed
upon printing.
Current make_pot.sh expects XSI behaviour and indeed e.g.
dash’s echo builtin implements this. echo builtins of other common
shells such as bash however, do not (by default). Avoid this portability
pitfall by just using printf at all relevant places. See:
https://pubs.opengroup.org/onlinepubs/9699919799/utilities/echo.htmlhttps://pubs.opengroup.org/onlinepubs/9699919799/utilities/printf.html
This also allows us to stop substituting each backslash with four
backslashes before piping to maybe_append. This substituion existed
to safely pass backslahes through maybe_append’s two layers of echo.
Also improve quoting. For consistency prefer single-quote quoting over
shell backslash escapes. QUote the unguarded *.lua to ensure it will
continue to work if there happen to be any lua files in the scripts
working directory.
This supersedes commits from wangqr/Aegisub which adjusted quoting and
echo usage to work with different shells. This is more throughout and
doesn't introduce semi-broken intermediate states.
0fbcaea8714aee271d03940181c7bc
(Note the second commit actually didn't manage to achieve the intended
portability, since several echo usages remained and it forgot to adjust
the backslash substitution, which was then fixed in the third commit.)
gettext can deal with AppData (and other XML files)
since version 0.19.6 (2015). We already depend on gettext anyway so
let’s utilise this, to get rid of two build dependencies and simplify
the extraction logic.
This depends on the removal of underscore prefixes
as done in a prior commit.
After this the developer name "Aegisub Group"
is now also treated as a translateable string.
gettext understands desktop files since 0.19 (2014),
so there’s no need for custom grep logic. This change
matches meson’s i18n which also uses gettext for this.
As a side efffect of this some long strings now get
broken into multiple lines (cosmetic only) and some
blank line gaps are normalised by gettext.
Cherry picks: 0e29c8d0e4
With fixups from: e98b7292f1
find may output matching files in any order.
To avoid unnecessary changes between regeneration
sort its output. For C++ sources xgettext’s --sort-by-file
already has an equivalent effect. All other files are already
statically listed in a fixed order.
Adopts part of: 4aee271d03
And document which aprts are still missing.
The keyword list is are copied from make_pot.sh.
Sorting by file helps keeping future PO(T) update diffs small'ish
and is also done in wangqr’s PO-file update procedure. (POT file
generation also sorts, but by preapplying `LC_ALL=C sort` to the
file list instead of using xgettext’s sorting option.)
Without this, meson’s aegisub-pot missed the majority of translatable
strings from C++ sources. While this brings the target much closerto
make_pot.sh, the remaining strings aren't straigthforward to integrate
unfortunately. At the time of writing meson has no support for iss files
and make_pot.sh uses custom logic to select only specific strings
from Lua and JSON files.
Even after this commit, POT updates should
thus still continue to use make_pot.sh.
It used to be genrated and installed by the old custom build system,
but apparently got lost in switching to meson.
Rename to .in.in instead of .template.in, since i18n otherwise gets
utterly confused failing to recognise the file type and both erroring
out with "no ITS rules found" during merge_file and failing to extract
any translatable strings.
Also drop underscores which were previously processed by intltool.
This doesn't yet automatically replace the release version and date
in the appdata file, but neither did the old build.
Note: while this slightly improves the situation, meson’s aegisub-pot
target is still missing many strings. make_pot.sh should be used.
meson’s i18n module get’s confused by .template and fails
to extract translatable strings, thus rename to in.in.
Note: while this slightly improves the situation, meson’s aegisub-pot
target is still missing many strings. make_pot.sh should be used.
a91c4f70c3 switched processing of the
desktop file from intltool to gettext-backed i18n.merge_file and dropped
leading underscores which used to be processed by intltool.
However, this commit omitted dropping the underscore from Keywords,
resulting in this field being stripped from the installed version.
Furthermore, make_pot.sh was broken since it too relied on the
underscore meaning POT regeneration will now remove the desktop files
strings from translation files.
Since it was not added to po/POTFILES meson’s native aegisub-pot target
also doesn't add desktop file strings to the translation pool. Though,
this target still misses a lot of other strings as well and shouldn't be
used yet anyway.
This commit removes leftover underscores and reintegrates desktop file
strings into both make_pot.sh. Integration into aegisub-pot is done in a
follow-up commit.
This consists of a meson option that
- sets a flag to make Aegisub read paths relative to the executable
- makes meson set up the symlinks like AppRun and .DirIcon as following
the AppDir format
and a CI workflow that builds with this option, installs to an AppDir
directory, and bundles it as an AppImage.
Modified from arch1t3cht/Aegisub#12 .
Co-authored-by: Fred Brennan <copypaste@kittens.ph>
wx doesn't seem to like the dialogs being created on some other worker
thread, which makes file dialogs opened by lua scripts crash in various
ways on Linux. Doing everything on the main thread hopefully fixes this.
FixesTypesettingTools/Aegisub#51 .
Add a utility library that wraps LWLibavSource and can parse its .lwi
file to obtain timecodes and keyframes. It also contains a function to
generate and save keyframes using WWXD or Scxvid. Update the default
scripts to use these functions.
The patched meson.build for dav1d just changes the include_directories
from 'include/dav1d' to 'include' and, while we're at it, removes the
pkg_config stuff to avoid a deprecation warning.
AV1 decoding is pretty broken in upstream ffms2, so point the wrap to a
fork that has some hack patches for it. BestSource works fine out of the
box, though.
This is nothing more than duct tape, do not pull.
In the long run boost::filesystem should just be replaced with
std::filesystem (as done in tgoyne/Aegisub:cmake), but this would
complicate the history and cause merge conflicts for a bunch of pulls.
Hence this horrible temporary solution.
When Avisynth is not installed or not functional, this would otherwise
cause a crash when trying to initialize Avisynth more than once, since
after the first time the refcount would have been incrased anyway.
Also, allow negative fsp.
This doesn't change the behaviour of the SpinCtrl increment/decrement
buttons, just the precision that can manually be set. Small values of
fsp can be useful as a hack to disable all ligatures. Precise values of
fscx/fscy can be useful to compensate for anamorphic resolutions. The
other fields were made more precise for consistency.
This fixes a crash on Windows when double-clicking the draggable
separator between the column headers "Command" and "Description" in the
hotkey configuration dialog.
This makes the internal juggling of fold data even more complicated, but
it has a number of advantages:
- Folds will preserved even when opening the file with Aegisub builds
that don't know about folding.
- Folds will be preserved by automation scripts that re-insert every
line.
- Folds do not break when adding or deleting lines in a text editor or
some other editor, as long as the lines involving folds aren't
touched.
- In particular, merging changes using tools like git will not break
folds.
- Copy/pasting folds also keeps the folds to some extent. This is
impossible to solve in full generality, but simple cases like
cut/pasting a folded section somewhere else work, as long as the file
isn't saved in between.
- This will give automation scripts full access to folding without
needing to set up any additional API functions.
Unlike the other arguments for the resize function, "range_in" does not
use the same format as in the frame props. A frame prop _Range=1 means
limited, while an argument range_in=1 means full range.
The highlighting distinguishes drawing commands from coordinates, and
colors x and y coordinates in different colors to make coordinates
easier to visually parse. Furthermore, in cubic Bezier curves, it
underlines the coordinates which corresponds to endpoints of the curves.
Aegisub will automatically override the YCbCr Matrix field of the
current file's Script Properties with the video's reported color space.
The FFMS2 provider guesses a color space for all videos, but we don't
do this for Vapoursynth. Thus, we now disable this overriding whenever
the colorspace isn't known.
On Windows, Python changes the application's locale upon being called,
which will break wxwidgets, causing various assertion error dialogs or
even crashes (for example when interacting with any sort of float edit
control). Saving the locale and restoring it afterwards seems to be
the only really possible way to remedy this.
The percent values used for the overscan masks follow the BBC's
guidelines, as in
https://en.wikipedia.org/wiki/Overscan#Overscan_amounts .
However, these measure the per-side width as opposed to the total
percentage of width/height being cut off. Thus, they should not be
divided by two when drawing the mask.
The new formula is just the inverse function of the CFR part of
the TimeAtFrame function.
To see how the previous implementation was faulty, see either the added
tests, or
- In Aegisub, open a dummy video with a frame rate of 23.976
- Make a subtitle event with start time 04:44.41
- Double-click the line to (supposedly) seek to its first frame
- This will seek one frame earlier than it should, and the event will
not be displayed on the resulting frame.
The secondary_editor is a wxTextCtrl, whose height is calculated from 2 rows of text. Using this height gives better consistency on screens with different DPIs, instead of using hard coded value like 50px
Since FileNotFound exceptions don't abort the provider search, opening
an invalid path would show errors such as "vapoursynth not found" when
VapourSynth wasn't installed, even if VapourSynth wasn't selected as the
default video provider.
This default script will be executed to load any file whose file name
extension is not .py or .vpy .
The gui code for setting the default script is still a bit wonky as it
doesn't fit the rest of the preferences pages nicely, but it works for
now.
Since FileNotFound exceptions don't abort the provider search, opening
an invalid path would show errors such as "avisynth not found" when
Avisynth wasn't installed, even if Avisynth wasn't selected as the video
provider.
When cancelling an automation macro from the progress dialog, the dialog
throws a UserCancelException. If the macro still runs to the end
afterwards (instead of calling aegisub.cancel or causing an exception),
the two return values are left on the stack. This causes assertion errors
due to check_stack when those are enabled.
Some of these got lost in the process before.
For now, default to 709 if the colorspace is unkown. More detailed
colorspace guessing can be added in the future.
These alignment flags would cause an assertion error
"Horizontal alignment flags are ignored in horizontal sizers"
when opening the dialog (mainly when importing styles from another
script).
As the assertion error says, the flags are ignored anyway, so they're
safe to remove.
After the audio provider rework, adjust the audio players to not
use the int16 mono downmixed audio unless necessary. Furthermore,
the pulseaudio-based player now controls the volume directly through
pulseaudio instead of by modifying the buffer. This also reduces latency
when changing the volume.
The entire set of GetAudio functions is quite messy now. After wangqr's
audio rework, it was split into GetAudio and GetInt16MonoAudio
functions, but now volume scaling is also necessary. Really, this should
go back to a type constructor based system with audio players being
allowed to choose what properties out of mono / 16 bytes / int samples /
volume they need.
These were giving false negatives on samba shares, which broke the font
collector. Windows also recommends to not use access checks in these
cases, and instead just see if the operations succeeds or not.
The file was mistakenly added when pulling wangqr's "time to video"
feature. I fixed the branch, but instead of merging the whole branch
again, let's just delete the file here.
Checking if the diff is '>' instead of '>=' will always fail if
tolerance=0, even if the colours are identical. This will cause the line
to get a startime greter than its end time, which is not desireable.
Rounding errors and limits of floating type precision might still affect
the comparison. An additional sanity check before calculation is added
to ensure the selected position and colour match within tolerance.
This allows us to refactor the search code to never check the starting
frame and guanrantees valid timings with start<end even with
rounding errors.
No other tooltips end with points, even if they're a sentence.These
ones aren't even sentences.
Also updates all *.po files accordingly, with an hacked in
POT-Creation-Date due to issues with make_pot.sh.
Float edits with a spinner are by default sized to fit their full
min-max range of possible values, so this makes min and max default to 0
and 100 (like it'd done internally in wx) instead of -DOUBLE_MAX and
DOUBLE_MAX.
Note that this does change the behavior of lua dialogs, but does not
contradict existing documentation or specification. It should only
affect scripts who either disobey the specification by specifying only
one value out of max/min, or scripts displaying these large float edits
by specifing a step, but no max or min.
Starting with AVISYNTH_INTERFACE_VERSION=5, this is how script
environments should be deleted. The previous code was causing crashes
when unloading AviSynth in certain scenarios, such as when failing to
open a file due to an incorrect path.
- DS2 Player has a similar structure to XAudio, so I don't see any reason why not to enable 1 channel+ audio.
- Haven't tried with 5.1 channel sources but I believe it should be the same as 2 channel (As in, Directsound will downmix the audio to 2 channel)
- Moved the volume setting to using the player directly and from some quick audio tests, -10000 is too soft. I tried with -5000 instead which seems to be alright.
- To allow for XAudio2 to work properly, we need to rework how does provider work since they only are used to be able to take in mono audio.
- Other providers have been dumbed down to accept single channel audio since originally aegisub only accepted 1 channel audio.
- meson.build has been modified to accommodate for xaudio, as we currently don't accept redistributable forms of xaudio, we need to work around the WinNT version.
- There has been 1 more fix to res.rc to allow for compiling on non tagged releases.
Checking if the diff is '>' instead of '>=' will always fail if
tolerance=0, even if the colours are identical. This will cause the line
to get a startime greter than its end time, which is not desireable.
Rounding errors and limits of floating type precision might still affect
the comparison. An additional sanity check before calculation is added
to ensure the selected position and colour match within tolerance.
This allows us to refactor the search code to never check the starting
frame and guanrantees valid timings with start<end even with
rounding errors.
No other tooltips end with points, even if they're a sentence.These
ones aren't even sentences.
Also updates all *.po files accordingly, with an hacked in
POT-Creation-Date due to issues with make_pot.sh.
Putting this logic for delaying changes in the TextSelectionController
isn't the cleanest, but all attempts at saving this state somewhere in
the Lua API instead turned out even worse. Also, the logic for inverted
selections probably does belong in there.
This way, automation scripts won't nuke the fold data on any lines they
modify.
This is done as userdata for now, so scripts that don't know what
they're doing don't break the folds even further. If scripts get more
access to folds, it should be with a safer and simpler API.
Line folds are managed using metadata of AssDialogue elements, and
saved in the project properties. They only affect the appearance of the
subtitle grid, and have no impact on other line operations.
The previous logic wouldn't compile on linux using clang.
With the meson build requiring c++14, there's no need for platform
specific code. Really, the entire agi::util::sleep_for function could be
removed entirely, but I'll keep the patch minimal for now.
This cuts down on the data duplication for the command names and help
texts, at the expense of copying a bit of code from the general toolbar
logic.
Modifying the toolbar Populate() function to also add additional buttons
appears very hard or impossible.
Meson 0.60.0 adds support for an "iconv" dependency that always does the
right thing for you, so that you don't need to think about it. It also
has some side advantages:
- you can do dependency fallback, and --force-fallback-for=iconv works
- it logs one line, not two, and that is "dependency found? yes or no"
Since Aegisub doesn't mandate the use of such new versions of Meson, we
cannot assume the dependency works. Instead, adapt to the version of
Meson being used: on new enough versions of Meson, use the new
dependency, but on older versions of Meson, use the pre-existing logic,
which isn't as nice but has been producing correct results so far.
- This refers to AviSynth < 2.5.6, which is from before 2005
- With the current setup (using GetAVSLinkage, these versions wouldn't
work anyway)
- The latest AviSynthPlus is bundled with the installer anyway
- Build the latest AviSynth headers using CMake, as proposed in #134
- Add and initialize AVS_Linkage to fix video loading
Video and audio playback now works in most cases, but still crashes for
some files.
Since ASS rendering depends on the storage size of the video libass
needs to know about it to render the subtitles correctly. If it isn't
told about the storage size libass uses the value from PlayRes{X,Y} as
a guess, but this isn't always correct.
With Aegisub currently always rendering at storage resolution
this ends up the same as the frame size.
This reverts commit 3d525d9938790211b02a31a16d716498957891fc.
The default hotkey is not required. People can just add a custom one
from the hotkey settings if they want this.
I'm not entirely sure, that this is how it's supposed to be added.
This requires editing the already existing hotkey.json file (shows path
in log, if it exists).
Should this also be added to hotkey.cpp?
The preview window would change in width, when the video was being zoomed.
Possibly, because on windows, zooming only calls UpdateSize once, but on
linux, it gets called multiple, like 8 times.
Rounding fixed it.
It is not picked up by the existing dependency() check because its
version is unknown to Meson, which fails the version constraint.
Besides, dependency('openal') seems to try "openal.framework",
which may or may not work on case-sensitive file systems.
Don't search for the dependency and set up fallback
only to throw it away if the option is disabled.
This is a waste, and this gives the user the illusion
that the request to disable the feature was ignored.
* [mac/build] update build steps
* [macos] bundle app
* [macos] build dmg
* [ci] setup macOS CI
* [i18n] TODO: rm WX locale files
* [deps] set main branch to main
* Fix osx-fix-libs.py
Create symbolic links, to make libicu happy
Rewrite the script in python3, as python2 is deprecated
* Add write permission before install_name_tool when doing osx-bundle
Fixwangqr/Aegisub#39
* Handle @loader_path in libboost on macOS
See wangqr/Aegisub#39
* [tools/mac] use python3
* [ci/mac] install & using system deps
* [ci/win] don't build fribidi:docs
* [ci/mac] trying openal-soft
* [ci/mac] use pulseaudio in CI
* [ci/win] only run aeg's test
* [ci/win] fix CI
* [ci/win] fix CI: not use dict `{}`
* [ci] run ci
* [ci/win] don't build docs
* [ci/win] remove args tail newline
* [ci/win] false->disabled
* Use md title format
Co-authored-by: Ryan Lucia <ryan@luciaonline.net>
* Recover file permissions.
* [ci/win] disable fontconfig
Co-Authored-By: Ryan Lucia <ryan@luciaonline.net>
* [ci/win] disable libass:fontconfig
Co-authored-by: wangqr <wangqr@wangqr.tk>
Co-authored-by: Ryan Lucia <ryan@luciaonline.net>
* [git] ignore IDE setting, Meson subprojects
* [win_installer] rename *.mo -> *.gmo
* [win_installer] only add aeg translations
* [win_installer] split aeg/wx translations
* [win_installer] gen & pack translations
* [git] ignore hunspell source dir
* [git] ignore uchardet source dir
* [test] Generate the test executable
run with `meson test`
* [test] add tools to remove test data files
`unset.bat FULL_PATH`
* [test] set to correct Error type
* [test] Set Source character sets to UTF-8 for MSVC
* [test] move test data files to build_root
* [meson/wxWidgets] fix warning: deprecated feature `cmake_options`
* [meson] fix warning: msvc does not support C++11
* [meson/fribidi] add original meson.build file
* [meson/fribidi] merge pr-151
https://github.com/fribidi/fribidi/pull/151
"meson: add fribidi_static_cargs to extra_cflags"
* [meson/fribidi] fix DEPRECATION
"Library fribidi was passed to the "libraries" keyword argument of a previous call to generate() method instead of first positional argument."
* [git] rm IDE dir
* [meson] remove old flags
* [meson] fixed in upstream
* [git] ignore IDE setting, Meson subprojects
* [win_installer] rename *.mo -> *.gmo
* [win_installer] only add aeg translations
* [win_installer] split aeg/wx translations
* [win_installer] gen & pack translations
* [git] ignore hunspell source dir
* [git] ignore uchardet source dir
* [test] Generate the test executable
run with `meson test`
* [test] add tools to remove test data files
`unset.bat FULL_PATH`
* [test] set to correct Error type
* [test] Set Source character sets to UTF-8 for MSVC
* [test] move test data files to build_root
* [git] remove IDE dir
* [git] ignore gtest subprojects
* [ci] run test in ci
* [meson/test] use more meaningful name
* [test] Add more comments and help msg
This option makes the dependency() call resolve to the subproject
which breaks the Lua 5.2 compatibility check.
This change just ignores the result of the dependency() call
if it resolves to the subproject and then re-fetches it
with a subproject() call later.
Alternatively we could explicitly handle the case where dependency()
resolves to the subproject, but that's just extra code for no
observable difference in behaviour.
Meson thought that a ARM64 cpu is aarch64, while
luajit only provided a file for "arm64". Since there is
no difference but the name between aarch64 and arm64, I
choose to bind aarch64 -> arm64.
Signed-off-by: Yi-Jyun Pan <pan93412@gmail.com>
Meson port instead of using CMake as I ran into issues with the
src directory (where uchardet.h is located) not being appended
to the include path, and on Windows I ran into a Meson issue
where a -D macro definition was being interpreted as a filename.
In the end a Meson port seemed simpler than working out the CMake
issues, as the CMakeLists.txt files were straightforward and easy
to port.
Note that because of the directory structure of the uchardet source
I had to change the include directive from uchardet/uchardet.h
to just uchardet.h. This is actually more in line with the uchardet
pkg-config file, which appends /usr/include/uchardet to the include path.
Previously, margins were clamped to 0..9999, but negative margins are
well supported by most renderers. In addition, previously lua
automations automations were able to produce these negative margin
values, and they would be saved correctly. However, re-opening the file
would clamp the values, and they could not be edited in the edit box.
This commit changes the clamping to be -9999..99999, and allows entering
(and editing) negative values in all relevant fields. In addition, this
makes the subtitle edit box margin fields take 5 characters instead of 4
to accommodate negative numbers up to 9999 (also the reason for raising
the upper bound).
I haven't checked whether reverting this breaks IME input, and if it doesn't what changed on wx's end. However, this is the commit that uses private symbols, and so reverting it lets us build against upstream wx. Even if this is a loss in functionality, for now it's fine.
System was made header-only by 1.74, so we can't check for it in the submodule, but when linking with older system boosts we need to make sure it gets linked in
monolithic build only and disabled precompiled headers due to limited cmake support in meson
still missing linker dependencies refs to Windows SDK libs
fixes an issue where the windows version of the git version update script generated the header in a place where it couldn't be found by the include in version.cpp
Using that makes use of gettext (and msgfmt) rather than intltool,
which is the proper way but requires a more recent gettext version and
I'm not comfortable adding that dependency for now.
Specifically, it would require updating the autotools chain as well
as the template files are incompatible.
See https://wiki.gnome.org/MigratingFromIntltoolToGettext for more info.
– Selection of different frequency mappings for the spectrum display,
from linear (current default) to logarithmic.
– Field added in the preference page to select the frequency mapping
– New preference key to remember this setting:
Audio/Renderer/Spectrum/FreqCurve
– Consistent display with high sampling rates (> 48 kHz)
– Fixed time shift with high quality settings
* Look for both system fonts and user fonts on Windows
* Move repeated font registry reading code to separate function
* Pass the files vector to read_fonts_from_key as a reference
* Allocate a larger filename buffer when necessary while reading fonts from registry
uchardet will report that a file is "ascii" if the first page has no bytes
>127, so we need to actually look at the entire file in case the first higher
byte is later in the file.
In 888be0607f some changes have been made to video_display.cpp which causes infinite recursion: the video display enables video/tool/cross tool when running VideoDisplay::Render() for the first time.
But when setting up the tool, the video box size is changed, which calls VideoDisplay::Render() again. So we need to set the tool BEFORE it modifies the box by "AddSeparator".
Fixwangqr/Aegisub#21
Co-authored-by: Charlie Jiang <cqjjjzr@126.com>
The thes file was using \r\n line ends, and \r was not stripped, causing the encoding string has a \r at its end.
Use binary mode and always use \n as line ends.
Previously visualSubToolBar has boarder while visualToolBar does not. So the width of the toolbar will change when toggling visualSubToolBar.
Now we remove the boarder so they have the same width. A separator is added at the top of visualToolBar instead to provide visual cue.
Fixwangqr/Aegisub#11
Instead of 1/3 screen per tick. So the distance of scroll no longer depends on the width of the audio box. Besides, 1/3 is feels to far for me when the audio box is wide.
See wangqr/Aegisub#5
The default fontsize is changed to 48. While it still looks small
on the default 720p video, it is actually bigger than
20 / 480 * 720 = 30
Here 720p is the default video resolution after 837d5a41d7. Some
common resolutions are also added to the preset list.
Fixwangqr/Aegisub#3
Aegisub crashes immediately after selecting any language
from the end of the list (above the 100th position).
This is because it can support no more than 100 languages.
This patch extends this limit up to 1000 languages (locales).
Fixes#131
Two notable changes: attempting to forcibly link icu4c no longer works for :reasons:, so you have to manually add it to the pkg-config and compiler/linker search directories. Additionally, homebrew removed build options from most packages as part of https://github.com/Homebrew/homebrew-core/issues/31510, and this includes luajit and boost. The former builds with gc64 by default on HEAD, and the latter with icu4c support by default, so this actually somewhat simplifies things.
Worth noting that this does NOT currently build, thanks to a few lines in libass that are not compatible with vc++. I'll bump the libass commit when MrSmile's pr gets merged and this is remedied. You can apply this commit to remedy that: 5317a8dd89
Also worth noting that while libass switched to nasm, I haven't done so here because it wasn't worth the effort, so I just made a questionable change to the yasm.targets to emulate nasm behavior
$(subst .c,.o,...) replaces '.c' with '.o' everywhere in pathnames. For
example, renaming the "Aegisub" folder to "Aegisub.cool" will make the
build system generate "Aegisub.oool/.../.o" objects.
https://www.gnu.org/software/make/manual/make.html#Text-Functions
FFMS2 originally would automatically attempt to guess the color space
if not manually specified, but this was removed in
7c2e08109d
and Aegisub never adapted to the change
VS 2015 has made distributing the dynamic CRT awful; the installer is
larger than Aegisub itself and app-local deployment requires an absurd
number of DLLs.
This deliberately doesn't perform any font substitutions as the sort of
people that use libass on Windows tend to perfer to manually pick
fonts with the correctly glyphs.
This significantly improves performance by avoiding the need to
construct std::strings of the things to be hashed, eliminating a bunch
of memory allocations and copies.
Supply a definition of socket.gettime() to remove the dependency on
LuaSocket, as installing binary deps from luarocks doesn't work with a
custom build of luajit.
We do need to check if the inactive markers are in the active set when
ctrl-dragging, as otherwise there'll always be a marker 0 pixels away to
snap to. Fortunately when ctrl-dragging all of the the markers involved
are by definition very close together, so it would be very difficult to
have enough markers to check for this to be a performance issue.
Closes#1823.
libass doesn't free any font data it requests until the ass_renderer is
destroyed, so recreate it every time the font is changed to avoid
holding on to extremely large amounts of data when the user quickly
switches between fonts.
Closes#1320.
boost::split_iterator type-erases the predicate, which makes it require
a virtual call per character (!) along with a heap allocation. As it
turns out we only ever need one predicate (comparing to a single
character), so replace it with a split_iterator that just does that.
MSVC 2013 and earlier have the _snprintf() function fill fields
completely, eliding the NUL character, if the printed string is
as long as or longer than the given buffer size. This is not the
case in POSIX snprintf(), it always ensures the string is terminated.
This causes several fields in GSI to become truncated and contain
NUL characters which should not appear there.
Both size_t and unsigned long long are aliases for
unsigned __int64 on Win64, and causes a duplicate definition.
Replacing all the integer type_name definitions with stdint.h
names should avoid this problem.
Has no real effect, it's what VS 2013 already uses and detailed
logging reveals that the build system doesn't even recognize v4
and assumes v12 anyway.
Might prevent builds with outdated tools.
Adds some additional MSBuild magic that chops dir names
from a given base, to avoid making deep, thin hierarchies
of dirs when building especially FFmpeg and libaegisub.
By not having it blindly run before the ClCompile target
it won't be checked every time you choose to build a single
file from the IDE. Instead it becomes a separate build stage
of its own.
Rasterized using Inkscape. Fixes the incorrectly blended white fringes
the old icon had at all resolutions below 512x512.
Also includes retina-sized images.
Fixes collection of ITC Cheltenham, ITC Tiffany, Zurich, Delicious,
Jacoby, Segoe WP, Agilita LT, and a bunch of dumb improperly modified
fonts I have lying around.
Closes#1806.
For whatever reason FcFontSetMatch sometimes returns something other
than the first result of FcFontSetSort, and every time they different
either FcFontSetMatch is wrong or they're both wrong.
Fixes collection of Adobe Jenson Pro and Arno Pro.
There have been a bunch of recurring problems with things not handling
zero dialogue lines or zero styles very well, so add one to the file on
commit if needed.
Closes#1814.
This is faster in all cases except for repainting a very small part of
the display with a very large number of range transitions and simplifies
the code a bit.
It hasn't actually been used for anything for a while, and if a single
normal-priority thread can make your system unresponsive it's time to
upgrade to an OS that's isn't garbage.
Not only is it unnecessary, but it actively breaks the mapping of
user-friendly encoding names in the Export dialog to the real encoding
names.
Closes#1789.
SubsTextEditCtrl::SetTextTo needs to update the selection via the text
selection controller or the text selection controller will have an
outdated cached state until the UpdateUI event is processed, which
soemtimes resulted in it not actually setting the selection when it
needed to be.
Freetype expects its IO to perform like mmap (i.e. seeking to be free),
but the non-unix implementation uses fseek and fread. This
implementation simply reads the entire file into a buffer, as fonts
generally aren't very big.
Cuts the time to recreate the fontconfig cache on my machine by about
60%.
Pass the paths to open by value to avoid ending up with a dangling
reference to a path stored in the MRU list which has been invalidated by
adding or removing items.
Closes#1777.
Keep track of which lines are displayed on the current video frame so
that video seeks only have to repaint the grid when the set actually
changes. Speeds up video playback on non-fbf stuff by ~15%.
At least on OS X, allocating the buffers is one of the more expensive
parts of video playback, and on an arbitrary 720p H.264 file with simple
subtitles this cuts CPU usage while playing by about 30%.
If the frame number hasn't changed and none of the currently visible
lines have changed, there's no need to rerender the subtitles and
redisplay the frame. Mostly helps with very slow opengl implementations
where actually painting the video display is expensive, as the rendering
is done on a background thread anyway.
Not the perfect place for it by any means, but that's where the rest of
the application-global stuff is these days and it removed the need for
automation stuff to include main.h (and thus all the wxApp garbage).
Opening the style list dropdown and switching between values with the
arrow keys doesn't update the value returned from ctrl->GetValue() on
OS X, but the event contains the correct value so use that instread.
The relative order of the on-file-load logic matters, so signals don't
work very well for it. Doing it all explicitly in Project fixes a bunch
of inconsistent issues related to restoring saved state.
Closes#1759.
A lot of the wxDialog subclasses don't actually override any virtual
functions, so there's no particular need for them to be subclasses at
all, and wxDialog's vtable is so huge that they actually contribute
measureable to the size of the executable.
It's modestly faster, significantly more type-safe, and doesn't assert
when there's too few arguments, which causes problems for plural forms.
Closes#1733.
It drags in a ton of crap, is kinda slow at runtime due to using
std::deque, and doesn't actually make the code much simpler than just
using a manual ring buffer.
Use std::chrono since it's a nicer API. Use a separate lock for playback
position so that the GUI thread isn't blocked for hundreds of ms while
snd_pcm_drain is waiting, and update the playback position after
decoding audio rather than before to avoid it being significantly wrong
when not using a cache.
If video was previously open and the user chooses to resample the script
for the new video, VideoController would try to update the subs for the
now-deleted old provider.
Fixes errors when trying to automatically open audio from dummy video
and improves UX a little by disabling the Open Audio From Video menu
item when there are no audio tracks.
Most of the dialogs in Aegisub have a public interface consisting of a
single function, so there's really no need to expose the actual dialog
classes to the rest of the program. Add dialogs.h with the declarations
of all of those functions (most of which are just ShowFooDialog()), and
kill a million other headers.
Remove it from the script info section and put it in its own section
that isn't tracked by undo and make it not stringly typed. Removes the
need for the gross hack where changes are slipped in just before saving
to circumvent the undo system, cuts down on the uses of string literals
to identify fields, and probably improves performance a little.
Add a new Project class which is responsible for everything related to
opening and closing audio, video, subtitles, timecodes and keyframes.
This pulls almost everything not directly related to playing audio/video
out of the audio and video controllers, pulls more crap out of
FrameMain, and happens to make things a little simpler in the process.
It is off by default on Windows due to that it breaks if a wx-using
library is loaded with LoadLibrary on XP, but that is not a use-case we
care about.
The shift from margins should be applied only to the line's position,
not to each point in the drawing, and the scale needs to compensate for
that the scale on the style is being changed as well.
Add modes where the borders to add or remove are automatically
calculated rather than forcing the user to do it manually, and hopefully
make it a bit clearer what exactly will happen.
wx comes very close to just making it work automatically, but it doesn't
translate the origin to the top-right corner and wxBufferedPaintDC
doesn't handle RTL at all, so reimplement it with some hacks.
Closes#1354.
Some of the toolbar buttons are seriously wrong when mirrored and there
doesn't seem to be a way to disable mirroring for the images while using
RTL layout for the toolbar itself, so double-mirror them so that they
end up with their original appearance.
Ideally the button images would be localizable as some of them actually
should be mirrored, but that's probably not worth the work involved.
See #1354.
This eliminates the need for a msys environment to build any part of
Aegisub. Also happens to make building FFmpeg much faster (~3 minutes
vs. ~10 minutes for me).
Eliminates the need to have a copy of git accessible in the msys
build environment and reduces the dependency on msys to just FFmpeg.
Probably trivially faster due to how slow shell scripts are on Windows.
This fixes crashes from touching dangling references to lines in the old
subtitle file in the handlers for the various signals triggered when a
new file is opened.
Sending events from background threads to the main thread is somewhat
expensive, so filter out progress updates which don't actually change
the progress on the background thread.
It seems that whitespace around the paths in the PATH, LIB, INCLUDE env
vars can break in some situations, making the build fail entirely.
Unfortunately this also makes the file harder to read and diff.
In preparation for switching to LuaJIT, which doesn't support PUC Lua's
thing of using C++ exceptions for lua_error.
Requires replacing all uses of lua_error (and things calling lua_error)
with custom versions that throw an exception instead and adding an
exception -> lua error wrapper at all C++ -> Lua boundaries.
This requires shuffling a bunch of stuff around to get a CLI lua
executable that uses an automation-like environment, but that's
something that'll be nice to have in the future anyway.
Busted indirectly depends on lfs and we currently can't use external
binary modules, so add a copy of lfs to the repo and build it with
Aegisub.
Use explicitly defaulted destructors for base cases when possible,
eliminate aggregate initialization constructors where possible, and push
some more stuff to NSDMIs.
Since luabins generates binary data which grows up to 3x by escaping,
it's more efficient to uuencode that instead.
A marker is placed as the first character of the value field, either
'e' for inline_string escaped text, or 'u' for uuencoded binary data.
The key is always inline_string escaped, as it will typically be
human readable.
See https://svn.boost.org/trac/boost/ticket/9360
It seems that the %X specifier is ignored if the input value is a char type
and the char is instead output verbatim and not in its integer value.
Casting it to an int works around this.
Results in the display seeking to the active line when audio is opened
(if seeking to active line is enabled), and initializes the scrollbar
properly.
Updates #1732.
Simply zero the memory for audio which hasn't been decoded yet, modify
the audio renderer to avoid caching blocks which aren't ready yet, and
add a progress indiciator to the audio display scrollbar.
ICU docs say not to create a new break iterator each time as
boost.locale does, and in fact creating the break iterator is about 90%
of the run time of the character counter, so use ICU directly and cache
the break iterator.
Only count characters outside of override blocks for the relative
positions to do the edits on each line, which handles the case where the
lines have different lengths of stuff in override blocks but the same
text.
Apply the modification at the cursor position to each selected line
rather than just setting all selected lines to the active line's text.
Still doesn't have good results in any nontrivial cases, but it's at
least sometimes useful.
ptr_vector hasn't been updated for C++11, so despite being specifically
designed to store pointers to objects it's less safe and not really any
easier to use than a regular vector of unique_ptrs
iterator_to requires that the object be in the list, which is sometimes
not the case when processing a commit which removed the active line or a
selected line. To handle this, add a checked version that returns
Events.end() when it is not in the list rather than crashing.
Make the vtables less absurdly huge (knocks 100KB off aegisub32.exe),
eliminate some copies of the values when constructing the options, and
use an enum class for the value type.
The grid needs to be able to map AssDialogue * to a row number, and just
storing it in the AssDialogue is nearly strictly better than a std::map
mapping events to rows. Probably will be of use elsewhere as well.
Register functions which create each type of provider rather than the
provider types themselves so that the concrete types don't need to be
publicly exposed, and use a static list of providers rather than
registering them at runtime.
The "previous" line for the first line was actually the last line of the
file, and the previous line for the second line was ignored entirely.
Closes#1719.
istringstream makes a copy of its input, which is not very nice for
performance. strstream doesn't do bounds checking, which is not very
nice for safety (and is deprecated).
Aegisub needs Lua built as C++ (for stack unwinding on errors). No one
packages it like that, so at the moment we can't actually use systme
packages for Lua.
Fixes a bunch of memory leaks on Automation errors.
LuaAssFile relied on the lifetime guarantees of intrusive lists (i.e.
you can pretty much do whatever the fuck you want and it'll work out),
but AssInfo objects are no longer owned by an intrusive list. Do awful
things to lazily copy the header section when needed to provide similar
lifetime guarantees.
They're only used by one thing (the drag visual tool), so calculating
them is pointless overhead most of the time, and it simplifies the
visual tool code much less than it complicates the grid code.
Store the data in vectors rather than AssFiles since even an intrusive
linked list has comically high memory overhead. Cuts memory usage of a
full undo stack with 15k lines by 65 MB for 32-bit and 130 MB for
64-bit. Also roughly halves how long it takes to copy the file for the
undo stack, and makes undo/redo a bit faster.
Despite being completely unmentioned in the documentation this appears
to be required to show up in the default programs UI on Windows 8.1.
Closes#1699.
These weren't merged from 2.1.9 with the rest of the file association
code, which results in the installer trying to reference nonexistent
resources and things breaking.
Updates #1699.
After shift-clicking on the grid the row to extend from doesn't equal
the active line, and clicking on the current active line to deselect all
other lines need to update the extend row.
When launching Aegisub from Finder (but not via open), using the UTF-8
version of the locale breaks things for whatever reason, but fortunately
it's unneccesary as paths on OS X are always UTF-8 even if the locale
isn't.
Closes#1685.
Color matching to dummy video makes zero sense, and setting it to None
results in breakage when someone opens dummy video for an already
typeset script.
Closes#1690.
--enable-stl replaces the implicit conversion to char/wchar_t* with one
to std::[w]string. This breaks conversions to boost::filesytem::path and
makes some ternaries ambiguous (which GCC helpfully reports as no
conversion existing rather than an ambiguous conversion).
Modifying the contents of a text control after IM processing happens in
the same cycle of the event loop seems to be seriously broken. Work
around this by disabling IM processing for time edits, as it should
never be relevant for them anyway.
Closes#1679. Closes#1680.
Don't delete the old copy of the file until after the new version is
committed to avoid a use-after-free when code tries to do things with
the active line in a COMMIT_NEW handler, as just trying to avoid doing
that hasn't really worked out.
The default Lua include path is encoded with the local charset rather
than UTF-8. This causes problems when the working directory includes
non-latin-1 characters which are valid in the local charset, as Aegisub
expects the string to be UTF-8. As the default path isn't very useful on
Windows (on other platforms it could potentially include things like
luarocks-installed modules), just clear it.
Closes#1676.
Switch to flat_map for trivially less memory usage and just use a global
map rather than a heap-allocated map since it's not
(de)initialization-order sensitive.
Eliminate the pointless runtime command name -> icon map and just make
commands responsible for their icons. Don't bother predecoding and
caching icons since it wastes some RAM (~10 MB) and decoding icons takes
a trivial amount of time.
Fix a few places where icons were being set for the wrong command names.
The standard MyThes dictionaries include words that we don't support
(nearly half the entries in the English dictionary have speaces in them)
which waste space, and they aren't UTF-8 which slows down loading as we
need to convert them.
Knocks a total of 10 MB off the thesaurus dictionaries.
Remove the faux-variadic make_unique since 2013 has real variadic
templates.
Switch to the non-deprecated windows version check functions.
Switch to the develop branch of boost since master doesn't fully support
VS 2013 atm.
Fix some assorted compilation errors.
Hardcode the paths to a bunch of Visual Studio things since it's now
expanding the relevant macros to nothing for whatever reason.
SetBitmap triggers a size event, which results in UpdateBitmap being
called again. For some reason, rather than being an infinite loop this
just resulted in the bitmap never being set.
libc++'s implementation of ostrstream seems to have incorrect
range-checking and does some out-of-bounds reads and writes whenever a
resize is needed, which results in crashes on 10.7 (but not 10.9 for
unknown reasons).
Lines need to be deselected before they're deleted, but new lines need
to be committed before they can be selected, so reorder things to
remove -> insert new line if needed -> commit -> update selection -> delete.
If the current frame is outside the bounds of the selected line, create
a duplicate of it at the current video frame. Otherwise, make the line
end on the previous frame and create a copy of it which starts on the
current frame (for split before).
For whatever reason this was sometimes being done after the commit,
which resulted in a lot of code only working by coincidence when the
grid commit handler happened to be the first one to be called.
It's quite rare to actually want a multi-frame line followed by a
single-frame line; in most cases they're either all a single frame (in
which case this behaves the same as it always has), or they're all
multi-line, in which case this should be more convenient.
The local docs it tries to open don't actually exists and if they're
ever added, the help contents command opens them anyway so I'm not sure
what the intended purpose of the command ever was.
If the script's matrix matches the video's matrix, use that even if
Force BT.601 is on, and if it's TV.601, use that even if Force BT.601 is
off.
This should mostly eliminate the common problems resulting from mixed
values of the Force BT.601 option.
Closes#1649.
For some reason the close event is getting sent multiple times and
them getting processed in the event loop for the confirmation dialog,
resulting in two dialogs.
Split int validating and double validating into two separate classes.
Make double parsing, validating and stringifying locale-aware. This is
far more complicated than it needs to be due to that Aegisub's locale
handling is a total mess.
Use DoubleValidator rather than wxFloatingPointValidator, because the
latter doesn't work with Aegisub's locale mess (on OS X it uses the C
locale for some things, and the locale reported by CoreFoundation for
others).
Closes#1568.
Passing nullptr to ostrstream's constructor is a nonstandard extension
that libc++ doesn't support, and as it turns out preallocating a buffer
doesn't actually meaningfully improve performance anyway.
Reverse iterators behave differently from regular iterators when
swapping (regular iterators continue to point at the same elements,
while reverse iterators do not), so instead use regular iterators with a
negative step for shifting down.
Showing zero is mildly misleading since it's actually default margin
rather than no margin, and this makes it much easier to spot the rows
where there's actually an interesting value.
Even without the animation the reported progress was an average of 1.5
blocks behind the actual progress, and with the animation it was two
blocks behind. Assuming constant-speed audio decoding it should now be
nearly accurate (it still ignores that the last block may be smaller
than the rest).
The wxLogger isn't created until it's needed by the main thread, so any
errors logged on a background thread before the logger was first used on
the main thread were simply being discarded. Fix this by forcing the
creation of the logger very early in the startup process.
If the text being pasted can be parsed as ASS dialogue events, then do
so even if the edit box is focused since it's rather unlikely that
the user actually wants ASS subtitle events in their subs.
Operate on characters rather than bytes in the dialog so that it
actually works with Kanji.
Rewrite the auto-matcher to handle more cases and add unit tests for it.
Unsurprisingly the munging around with the menu bar implementation
details has resulted in things breaking, but fortunately the problem it
was working around (#1314) appears to have been fixed entirely.
Closes#1628.
Rather than pinning the time at the center of the audio display, pin the
time at the mouse's position since that's the area the user probably
cares about. If the mouse is not over the audio display (such as if the
user is adjusting the zoom via the sliders rather than the mouse wheel),
just use the center as it used to.
The X scale/offset was being used for both X and Y coordinates.
Only non-negative integers were supported. xy-VSFilter and libass both
now support non-integer coordinates, and negative coordinates have of
course always been valid. Resampled coordinates are now rounded to
eighth-pixels rather than whole pixels.
Calling GetReverseUnicodePosition on the selection positioning after
calling SetText resulted in the conversions between UTF-8 indices and
wxString indices just being a very slow no-op.
wxEVT_CLOSE_WINDOW is only triggered from the platform's close buttons,
not cancel buttons/ESC, so modeless dialogs closed in that way were not
getting deleted.
Video frames aren't actually BGRA; the alpha channel is actually just
garbage since CSRI uses 0 for opaque and other things use 255. To work
around this, add a custom colorspace converter.
Fixes the subtitles preview in the style editor and copying frames to
the clipboard/saving them.
Closes#1621.
Poorly-written antivirus software briefly lock newly written files to
scan them for viruses, which makes the rename from the temp file to
actual file fail. Work around this by retrying the rename up to ten
times.
Closes#1620.
Removing the first character from the non-custom case resulted in an
error when the value was empty and resulted in nothing happening when a
valid value was present.
wxSpinCtrlDouble only updates its value when the spin button is clicked
or it loses focus, and in some cases we want to read the value while it
still has focus.
Closes#1610.
Some example uses:
-- ~special snowflake~ OK/Cancel
aegisub.dialog.display(config, {ok='Accept', cancel='Cancel'})
-- On OS X the 'Help' button will be just a left-aligned ?
aegisub.dialog.display(config, {ok='OK', cancel='Cancel', help='Help'})
-- Each button in its own subtable to preserve passed order
-- Unnecessary when using only IDed buttons since the passed order will
-- be ignored in favor of the platform-standard order
aegisub.dialog.display(config,
{{ok='Accept'}, {cancel='Cancel'}, {help='Help'}, 'Another Button'})
In some cases the passed labels will be ignored in favor of the
platform-standard labels.
Available IDs:
ok
yes
save
apply
close
no
cancel
help
context_help
Note that many combinations of button IDs do not make sense and may have
strange effects.
Buttons with an ID of 'cancel' return false, as if ESC was pressed. A
button with an ID of 'close' results in that button being triggered on
ESC rather than cancel.
Buttons with an ID of 'ok', 'yes' and 'save' are set as the default
affirmative button for the dialog.
Closes#1609.
The standard controls aren't particularly usable, but the OS X
implementation of wxDVC doesn't actually support custom renderers and
not very usable is mildly better than entirely nonfunctional.
Updates #1589.
0.11 has a bug that makes it crash on MoonScript, and 0.12 is much slower.
LPeg isn't packaged as a C library and consists of a whopping two files
so just compile it as part of Aegisub.
Set it to the first line not part of the selection after the selection
begins if there are any, and the last line remaining in the file if not
(i.e. the last line before the selection).
Closes#1595.
API is mostly unchanged other than the addition of a lot more flags.
Should be less buggy since it has an actual test suite, and generally
has a more powerful regex syntax with better support for Unicode.
The bindings are written in MoonScript. For now the compiled form is
store in the repo for convenince.
Use ICU to split the text into characters rather than assuming that one
wchar_t == one character in some places, and one char == one character
in other places.
This marginally increases memory use, but vastly speeds up pretty much
everything when a file has attachments (other than extracting the
attachments, but that's generally IO-bound anyway).
This is currently only correct if the UI language is the same as the
subtitles language as the global locale is used, but it should still
never be worse than the small hardcoded table of word splitters.
Closes#1206.
Relying on TextFileReader to do the charset detection results in the
user being prompted to pick a charset twice when it can't be
auto-detected, since the result from trying to open the subtitles as
timecodes was not being reused.
Closes#1512.
Do proper unicode case-folding for case-insensitive searching rather
than converting only ascii characters to lowercase. The Turkish 'i' is
still not handled correctly (since it's the only place where
case-folding is locale-dependent), but that's probably not worth caring
about as long as we don't have a Turkish UI translation.
This affects both the find/replace dialog and the select lines dialog.
Closes#1342.
Add SubsController, which deals with things like what subtitle file is
currently open, rather than the contents of the current subtitle file.
Move the rest of the relevant logic from FrameMain there in addition to
all of the stuff from AssFile.
Use boost::filesystem::path for all paths, and std::string for all other
strings, converting to/from wxString as close to the actual uses of wx
as possible.
Where possible, replace the uses of non-UI wxWidgets functionality with
the additions to the standard library in C++11, or the equivalents in
boost.
Move the path token management logic to libaegisub (and rewrite it in
the process).
Add a basic thread pool based on asio and std::thread to libaegisub.
This touches nearly every file in the project and a nontrivial amount of
code had to be rewritten entirely, so there's probably a lot of broken
stuff.
This seems to be what most users expect, rather than pasting downwards
from the active line, and is what pre-3.0 versions did as long as the
selection was continuous (since it pasted over from the first selected
line rather than the active line).
Emit a separate event for when a color is picked rather than horribly
overloading the onclick event, and switch to a wxButton base since
wxBitmapButton no longer does anything useful.
Decouple it from the subtitles grid and do everything through the
project context instead.
Actually obey the "In Selection" option for things other than Replace
All.
Cut down on the amount of logic duplicated between Replace All and
normal find/replace.
Start searching from the current cursor position, rather than the last
found position.
'isComment ? "" : actor' results in actor decaying to a const char * in
the local charset rather than converting "" to wxString, which breaks
things when unicode is involved.
wxString::Format is astonishingly slow, and while this normally doesn't
particularly matter, AssEntry::EntryData's speed can become relevant on
large files.
Switch to a slightly more manual approach that's about four times as
fast.
Nearly all of the files including main.h are doing so only for OPT_GET
and friends, which are rather unrelated to the main things that main.h
declares.
The video source's copy of the subtitle file was being discarded once it
was fully loaded (as it was previously never needed again), so amending
it only worked if the user had made a non-amend commit after the last
seek.
Aegisub's downmixer is terrible for anything with more than two channels
(averaging all the channels together is not even vaguely close to the
correct thing to do for 5.1), so libavresample should be far better.
Audio providers should be responsible for determining whether or not
they can open a URI, since it's not really possible for anything else
to. As such move the (lack of) parsing of dummy audio URIs to
DummyAudioProvider, and try to create DummyAudioProviders in
AudioProvider::CreateProvider.
Creating an indexer can fail for reasons other than the file not
existing. This check is still not completely correct, since FFMS2 uses
the wrong error codes in a bunch of places.
When the set of changed lines is populated, only copy those lines rather
than the entire file. On large files, this makes amend commits roughly
twice as fast when video is open.
Currently this is only populated when commits are amended, for the
simple reason that it's the only time that AssFile knows what lines
changed. It is probably worth expanding this in the future.
Always store the original string form and convert it to the desired type
on demand. In most cases this is significantly faster as typically only
a few of the parameters in a line are actually read. In addition, this
makes it so that parameters which did not actually change are always
reserialized in their original form.
Delete VariableData as it's no longer used by anything.
Closes#1513.
Now that all the other junk is being dropped from the file, script info
lines are the only thing left which was a concrete AssEntry, and
AssEntry wasn't actually a very good way to store them.
Character counts are a pretty terrible measure of anything, but it's
still the main measure of length used in most subtitling standards
(because subtitling standards are written under the assumption that the
tools are terrible (which is generally an accurate assumption)).
100 no-op non-amend commits on a subtitle file with 6689 dialogue lines,
with the undo limit set to 100:
Without flyweight:
No video open:
Initial memory usage: 30.6 MB
Final memory usage: 498.0 MB
Elapsed time: 6.3 seconds
Video open, using libass:
Initial memory usage: 54.3 MB
Final memory usage: 653.3 MB
Elapsed time: 23.7 seconds
With flyweight:
No video open:
Initial memory usage: 26.0 MB
Final memory usage: 104.5 MB
Elapsed time: 3.0 seconds
Video open, using libass:
Initial memory usage: 46.7 MB
Final memory usage: 251.8 MB
Elapsed time: 13.0 seconds
No video open:
Memory usage: -79%
Time: -52%
Video open:
Memory usage: -61.5%
Time: -45%
100 no-op amend commits on a line in the middle of a subtitle file with
6689 dialogue lines, with video open:
Without flyweight:
Initial memory usage: 48.2 MB
Final memory usage: 182.3 MB
Elapsed time: 22.3 seconds
With flyweight:
Initial memory usage: 39.8 MB
Final memory usage: 165.8 MB
Elapsed time: 13.8 seconds
Note: The large jump in memory usage here is due to that the benchmark
is blocking the main thread, so at the end there are ~100 video frames
waiting to be displayed.
This eliminates the need to explicitly delete the parsed blocks.
Also go ahead and remove AssDialogue::Blocks since it's really not a
very good idea in general.
FcFontRenderPrepare discards excess family names and fullnames, so in
some cases it may be impossible to verify that the font returned by it
is the one we asked for. To work around this, prefilter the available
fonts to only do fontconfig's matching on the ones with the correct
names.
Make two passes over the file to collect all of the styles before
processing the dialogue lines, as the styles section can legally come
after the dialogue lines which use those styles.
Dealing with multiple instances of each section makes the code
significantly more complicated, and in most cases Aegisub doesn't
actually bother to handle it correctly.
Previously, Aegisub wrote its "Script generated by" lines without a
semicolon to mark them as comments. As a result, it prepended these
lines every time a script was loaded and then saved again. Fix.
Also fix a bug that caused the ASS parser to treat comments before
the [Script Info] section as non-blank lines, therefore inserting a
duplicate section header whenever such comments were present.
Inserting lines already in an intrusive list at a different position
does not work, unlike with unintrusive lists. Instead, swap each item in
to the correct place.
There's really no reason not to include it even if the right-click
wasn't on a word, and leaving it out can lead to confusing situations
when the dictionary file for the user's currently selected language
doesn't exist.
This currently consists of just a list of autosave files and the
timestamps of the versions available for each file. It'd probably be a
good idea to add more information to this later, such as the original
file path and perhaps a preview of the file?
Closes#781.
1.6.0 removed support for 'make install' as the maintainers are
apparently unaware that their library is not actually a unique snowflake
and that having to use identical compilation settings for every library
is in fact a completely normal thing for C++ libraries that people
somehow manage to deal with.
Building the tests now requires passing the root directory of a copy of
the googletest source to make, via 'make GTEST_ROOT=~/path/to/gtest'.
The syntax highlighting isn't especially slow, but the stylesneeded
event seems to be sent repeatedly until the styles aren't changed, so it
was pointlessly reparsing and restyling the text on every idle event.
Pass the ASS parser a [Events] line before feeding it dialogue lines so
that the dialogue lines are parsed as such rather than as script info
lines.
Updates #1554.
At very large and very small sizes, setting the window's size with
SetSize may not actually result in the window changing to the requested
size. Once this happened future adjustments to the size were incorrect,
and the video display would sometimes not fit in the window, and on os x
the window could get set to a negative size, with weird results.
wxWidgets currently doesn't generate menu open or update events for the
Ubuntu menubar (since GTK doesn't send the relevant signals to the
application), which we rely on.
Updates #1531.
This makes it so that autosave files are (almost) never overwritten, so
that opening an old version of the file won't result in the autosaves
from a newer version being replaced.
Clean up old autosave files on startup to limit the total to 100
files/100 MB. Maybe make this configurable in the future?
Closes#1155.
wxRemoveFile notifies the user of errors via a dialog box, and errors
when cleaning the cache are normally just due to two copies of Aegisub
trying to clean it at the same time.
Mostly just bugs in unreachable code and stylistic things, but there's a
few incorrect reachable things that were working by coincidence and
gratuitous dynamic_casts.
The valid bounds for a selection are not updated until the contents of
the edit box are updated on commit, so set the selection after
committing rather than before.
It is very unlikely that anyone has ever actually used ASS2 for
anything, and in practice it was not usable anyway as the bottom margin
was always either ignored or blindly overwritten.
Instead, just pass the provider to the constructor and let the player
worry about when to create its things. The ability to explicitly open
and close the playback wasn't actually required for anything, and it
complicated the implementations of the players.
Actually use the limits set in the preferences dialog rather than always
limiting the MRU lists to 16.
Currently only the Find and Replace limits can be set to above 16 due to
how the open recent items commands are implemented.
Closes#1528.
This is a potentially sensible operation (to convert from VFR to/from
the video's average FPS), so there's no reason to require that the input
and output timecodes be from different places.
Closes#1530.
AegisubLocale is created extremely early in the application
initialization, and using StandardPaths in its constructor resulted in
?user being set to an incorrect value as the app name hadn't been set
yet.
Use wxTranslations directly rather than going through wxLocale. This
significantly simplifies the code, eliminates the hardcoded list of
languages for non-windows, and makes it possible to use mismatched
languages and locales.
Closes#1508.
When auto scrolling is enabled, scroll the audio display after a click
near an edge to mimic the effect of scrolling to show the new audio
selection (as 2.1.9 did), but hopefully without all the weird
undesirable cases that had.
wxLANGUAGE_ENGLISH corresponds to the "en" locale on Linux, which is
generally not installed, which resulted in English actually being the
system default locale.
Updates #1508.
Download release builds [here](https://github.com/arch1t3cht/Aegisub/releases), or the latest CI builds [here](https://github.com/arch1t3cht/Aegisub/actions?query=branch%3Afeature+event%3Apush).
The release page also has detailed list of all changes and new features. If you're interested in the technical details or want to compile yourself, read on.
### Don't we have enough Aegisub forks already??
We absolutely do, and I'm aware that adding another one [doesn't sound like](https://xkcd.com/927/) a [good idea on paper](https://cdn.discordapp.com/attachments/425357202963038208/1007103606421459004/unknown.png). However,
- None of the existing forks are completely satisfying at the moment:
- [wangqr's fork](https://github.com/wangqr/Aegisub) is actively maintained, but focussing more on stability. It's missing most of the modern features.
- [AegisubDC](https://github.com/Ristellise/AegisubDC) has the most modern features (in particular video-panning), but is Windows-only and not actively maintained anymore.
- [The TypesettingTools fork](https://github.com/TypesettingTools/Aegisub) is the one that will one day become the upstream version and builds relatively effortlessly on all operating systems, but at the moment it's not moving much. It's the base for this fork, and I hope to one day merge most of these additions into it.
- Only PR'ing the changes in here to various forks would cause even more chaos
- ~~I try to convince myself that this isn't really a "fork" in the traditional sense - one which aims to provide extended support and stability fixes. It's a collection of new feature additions which I built myself, together with some of the most important new features floating around other forks.~~ At this point it's probably too late to still be saying this. Still, the general mission hasn't changed. This fork collects new features and critical bugfixes, but won't be putting extra time into maintenance aspects like cleanup and refactors. Partly, this is also because any big refactors would make it harder to pull these changes into upstream repositories or future forks.
While this is usually also the version of Aegisub I'm currently using, I make no promises on stability. **Don't** use this version if you're just looking for any version of Aegisub - this is mostly intended for typesetting and other advanced usage.
### Organization
Being a collection of different feature additions, this repository consists of a set of branches for different features, so that they can easily be merged into other repositories. The [`feature`](https://github.com/arch1t3cht/Aegisub/tree/feature) branch merges together all the features I deem as currently usable. Due to the structure of the repository, I will be force-pushing to this branch and some of the individual branches very frequently, so they're not ideal for basing further branches on.
The `cibuilds` branch makes some CI builds of snapshots of `feature` at relevant points in time.
### Branch/Feature list
This list is for navigating the repository. Go to the [release page](https://github.com/arch1t3cht/Aegisub/releases) for a more structured changelog.
- [`folding`](https://github.com/arch1t3cht/Aegisub/tree/folding): Add the ability to visually group and collapse lines in the subtitle grid
- [`lua_api`](https://github.com/arch1t3cht/Aegisub/tree/lua_api): Add new functions to the Lua automation API, like controlling the selection or cursor in the text edit box
- [`vector_clip_actions`](https://github.com/arch1t3cht/Aegisub/tree/vector_clip_actions): Make the different modes of the vector clip tool (lines, bezier curves, adding points, etc) bindable to hotkeys
- [`color_picker_fix2`](https://github.com/arch1t3cht/Aegisub/tree/color_picker_fix2): Add an option (under "Interface") to restrict the color picker to the window, which fixes the color picker on Linux in a lot of cases.
- [`avisynth`](https://github.com/arch1t3cht/Aegisub/tree/avisynth): Reenable Avisynth support on Windows and enable Avisynth on Linux
- [`bestsource`](https://github.com/arch1t3cht/Aegisub/tree/bestsource): Add BestSource audio and video source. This source is slower than others by multiple orders of magnitude, but in exchange it can guarantee exact seeking.
- [`vapoursynth`](https://github.com/arch1t3cht/Aegisub/tree/vapoursynth): Add Vapoursynth audio and video source
- [`bugfixes`](https://github.com/arch1t3cht/Aegisub/tree/bugfixes): Various fixes necessary for compilation. Most branches are based on this.
- [`workarounds`](https://github.com/arch1t3cht/Aegisub/tree/workarounds): Same as `bugfixes`, but these are hacky fixes that probably shouldn't be pulled without more work.
- [`misc`](https://github.com/arch1t3cht/Aegisub/tree/misc): Other miscellaneous additions
- [`wangqr_gui`](https://github.com/arch1t3cht/Aegisub/tree/wangqr_gui): Merge wangqr's changes regarding the GUI. In particular, add high-DPI compatibility.
- [`misc_dc`](https://github.com/arch1t3cht/Aegisub/tree/misc_dc): Miscellaneous changes taken from AegisubDC
- [`xa2-ds`](https://github.com/arch1t3cht/Aegisub/tree/xa2-ds): Add XAudio2 backend and allow stereo playback for some other backends, by wangqr and Shinon.
- [`stereo`](https://github.com/arch1t3cht/Aegisub/tree/stereo): Add multi-channel support for the other audio backends where possible.
- [`video_panning_option`](https://github.com/arch1t3cht/Aegisub/tree/video_panning_option): Merge [moex3's video zoom and panning](https://github.com/TypesettingTools/Aegisub/pull/150), with several bugfixes and more options to control zoom behavior
- [`spectrum-frequency-mapping`](https://github.com/arch1t3cht/Aegisub/tree/spectrum-frequency-mapping): Merge EleonoreMizo's [spectrum display improvements](https://github.com/TypesettingTools/Aegisub/pull/94), and also make Shift+Scroll vertically zoom the audio display
- [`wangqr_time_video`](https://github.com/arch1t3cht/Aegisub/tree/wangqr_time_video): Merge wangqr's feature adding a tool for timing subtitles to changes in the video
### Troubleshooting
I'll gladly take any bug reports, but if you encounter an issue, please check first if it occurs only on my fork, or also on [earlier TSTools builds](https://github.com/TypesettingTools/Aegisub/actions).
If it wasn't introduced by my fork, I can still take a look, but I can't promise anything.
You can find me for support on various servers, including the cave and the TSTools server linked below.
#### Aegisub on Linux doesn't recognize my GTK theme
This is probably because you're building with wxgtk2. Building with wxgtk3 fixes this, but causes some problems of its own (notably the broken color picker, occasional crashes when opening file dialogs from automation scripts, and general layouting issues).
The exact way of switching depends on your Linux distribution, but essentially you need to ensure that `wx-config` or the next best variant of it points to wxgtk3. If it points to wxgtk2 by default and deinstalling wxgtk2 isn't an option, you can also temporarily move it out of the path or use a `native-file` in your meson project. Then, fully reconfigure meson using `meson configure --clearcache` and `meson setup --reconfigure`.
#### The video is desynced / Frames don't appear at the right time
This is probably due to the ffms2 seeking bug ([#394](https://github.com/FFMS/ffms2/issues/394)). On Windows, this specific regression shouldn't happen anymore. On Linux, you need to install the latest git version of ffms2 - for example the [`ffms2-git`](https://aur.archlinux.org/packages/ffms2-git) AUR package on Arch linux, or just compile it yourself.
If it's not because of this particular bug, you can also try an alternative video source like LSMASHSource via Avisynth or Vapoursynth, or BestSource.
#### On Windows: Aegisub crashes whenever I open a video
If you're compiling yourself, try adding `--force-fallback-for=zlib` to the meson options.
### Compilation
If you're just looking to install Aegisub, you might want to check out the [releases page](https://github.com/arch1t3cht/Aegisub/releases) or the [CI builds](https://github.com/arch1t3cht/Aegisub/actions) first.
For compilation on Windows, see the TSTools documentation below. Also check the [GitHub workflow](https://github.com/arch1t3cht/Aegisub/blob/cibuilds/.github/workflows/ci.yml) for the project arguments.
On Arch Linux, there is an AUR package called [aegisub-arch1t3cht-git](https://aur.archlinux.org/packages/aegisub-arch1t3cht-git). It's not maintained by me but seems to work.
On other Linux distributions or for manual compilation you can use this package or the [TSTools PKGBUILD](https://aur.archlinux.org/packages/aegisub-ttools-meson-git) as a reference, in particular for installing the necessary dependencies if you don't want to compile them yourself.
If all dependencies are installed:
- Install Meson
- Clone the repository
- In the repository, run `meson setup build --buildtype=release` for the default configuration. See below for further options.
- `cd` to the `build` directory and run `ninja`
- You'll get an `aegisub` binary in the `build` folder. To install it to a system-wide location, run `ninja install`. To install to `/usr` instead of `/usr/local`, pass `--prefix=/usr` when configuring or reconfiguring meson.
- When recompiling after pulling new commits, skip the `meson setup` setup and just immediately run `ninja` from the build directory - even when the build configuration changed.
#### Compilation flags
Some features are not enabled by default. To enable them, pass `-D<feature>=enabled` with the `meson setup` command:
- `-Davisynth=enabled`: Avisynth support
- `-Dbestsource=enabled`: BestSource
- `-Dvapoursynth=enabled`: Vapoursynth support
You can also disable options that are active by default in the same way. Check the file `meson_options.txt` for all options.
To change the options of an existing build directory, run `meson setup --reconfigure <new arguments>` from inside the `build` directory.
### Dependencies
Apart from the dependencies for the TSTools version, there are some additional dependencies. These are cloned and compiled from scratch if not found, but you might want to install binaries instead:
- `xxhash`: For BestSource
- `ffmpeg`: Becomes a direct dependency when compiling with BestSource
- `avisynth` (or `avisynthplus`): Optional run-time dependency for the Avisynth audio/video provider
- `vapoursynth`: Optional run-time dependency for the VapourSynth audio/video provider
The following VapourSynth plugins are used by the default scripts set in the default configuration:
- [`lsmas`](https://github.com/HomeOfAviSynthPlusEvolution/L-SMASH-Works): For LWLibavSource
- [`bs`](https://github.com/vapoursynth/bestsource): For BestSource, used for audio
- [`wwxd`](https://github.com/dubhater/vapoursynth-wwxd) and [`scxvid`](https://github.com/dubhater/vapoursynth-scxvid) (depending on settings): For keyframe generation
# Aegisub
For binaries and general information [see the homepage](http://www.aegisub.org).
The bug tracker can be found at https://github.com/Aegisub/Aegisub/issues.
Support is available on [Discord](https://discord.com/invite/AZaVyPr) or [IRC](irc://irc.rizon.net/aegisub).
## Building Aegisub
### Windows
Prerequisites:
1. Visual Studio (Community edition of any recent version is fine)
2. The June 2010 DirectX SDK (the final release before DirectSound was dropped)
3. Python 3
4. Meson
5. CMake
6. Powershell execution policy set to Unrestricted
There are a few optional dependencies that must be installed and on your PATH:
1. msgfmt, to build the translations
2. InnoSetup, to build the regular installer
3. 7zip, to build the regular installer
4. Moonscript, to build the regular installer
All other dependencies are either stored in the repository or are included as submodules.
2. From the Visual Studio "x64 Native Tools Command Prompt", generate the build directory: `meson build -Ddefault_library=static` (if building for release, add `--buildtype=release`)
3. Build with `cd build` and `ninja`
You should now have a binary: `aegisub.exe`.
Installer:
You can generate the installer with `ninja win-installer` after a successful build. This assumes a working internet connection and installation of the optional dependencies.
You can generate the portable zip with `ninja win-portable` after a successful build.
### OS X
A vaguely recent version of Xcode and the corresponding command-line tools are required.
For personal usage, you can use pip and homebrew to install almost all of Aegisub's dependencies:
From within the Moonscript repository, run `bin/moon bin/splat.moon -l moonscript moonscript/ > bin/moonscript.lua`.
Open the newly created `bin/moonscript.lua`, and within it make the following changes:
1. Prepend the final line of the file, `package.preload["moonscript"]()`, with a `return`, producing `return package.preload["moonscript"]()`.
2. Within the function at `package.preload['moonscript.base']`, remove references to `moon_loader`, `insert_loader`, and `remove_loader`. This means removing their declarations, definitions, and entries in the returned table.
3. Within the function at `package.preload['moonscript']`, remove the line `_with_0.insert_loader()`.
The file is now ready for use, to be placed in `automation/include` within the Aegisub repo.
## License
All files in this repository are licensed under various GPL-compatible BSD-style licenses; see LICENCE and the individual source files for more information.
The official Windows and OS X builds are GPLv2 due to including fftw3.
-- Copyright (c) 2012, Thomas Goyne <plorkyeran@aegisub.org>
--
-- Permission to use, copy, modify, and distribute this software for any
-- purpose with or without fee is hereby granted, provided that the above
-- copyright notice and this permission notice appear in all copies.
--
-- THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
-- WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
-- MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
-- ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
-- WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
-- ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
-- OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
--
-- $Id$
localtr=aegisub.gettext
script_name=tr"Clean Script Info"
script_description=tr"Removes all but the absolutely required fields from the Script Info section. You might want to run this on files that you plan to distribute in original form."
Text="The SubWCRev.exe program was not found. If you already have TortoiseSVN installed, override the SubWCRevExe property to the location of the program."
/>
<Error
Condition="!Exists('%(SubWCRev.Repository)')"
Text="The working copy was not found. Location specified: %(SubWCRev.Repository)"
/>
<Error
Condition="!Exists('%(SubWCRev.InputFile)')"
Text="The template file was not found. Location specified: %(SubWCRev.InputFile)"
Condition="'$(DirectXSDKDir)'=='' or !Exists('$(DirectXSDKDir)\Include\dsound.h')"
Text="DirectX SDK not found, check that the DXSDK_DIR environment is set. If you do not want to use DirectSound set the AegisubUseDSound property to false in a user config."
Description="Specify whether to use config_windows.h or the configuration on this page to control features. The settings on this page still control automatic compiler and link settings for the features."
/>
<BoolProperty
Name="AegisubUseAutomationLua"
Category="Features"
DisplayName="Use Automation 4 Lua"
Description="Specify whether to include Automation 4 support and the Lua engine. This is included and requires no configuration."
/>
<BoolProperty
Name="AegisubUseDSound"
Category="Features"
DisplayName="Use DirectSound"
Description="Specify whether to use the DirectSound-based audio player. This requires the DirectX SDK installed."
/>
<StringProperty
Subtype="folder"
Name="DirectXSDKPath"
Category="Paths"
DisplayName="DirectX SDK root path"
Description="Location of DirectX SDK installation for DirectSound support, this is usually auto-detected."
/>
<BoolProperty
Name="AegisubUseAvisynth"
Category="Features"
DisplayName="Use Avisynth"
Description="Specify whether to support Avisynth for video and audio providers. This is experimental on x64."
/>
<BoolProperty
Name="AegisubUseFfms"
Category="Features"
DisplayName="Use FFmpegSource2"
Description="Specify whether to support FFmpegSource2 for video and audio providers. You may need to configure the location of the library on the Library paths page."
/>
<StringProperty
Subtype="folder"
Name="FFmpegSourceIncludePath"
Category="Paths"
DisplayName="FFmpegSource2 include path"
Description="Location of FFmpegSource2 include files"
/>
<StringProperty
Subtype="folder"
Name="FFmpegSourceLibraryPath"
Category="Paths"
DisplayName="FFmpegSource2 library path"
Description="Location of FFmpegSource2 library files"
Description="Specify whether to use Freetype 2 for font enumeration in the Font Collector. You will need to configure the location of the Freetype 2 library on the Library paths page."
/>
<StringProperty
Subtype="folder"
Name="Ft2IncludePath"
Category="Paths"
DisplayName="Freetype 2 include path"
Description="Location of Freetype 2 include files"
/>
<StringProperty
Subtype="folder"
Name="Ft2LibraryPath"
Category="Paths"
DisplayName="Freetype 2 library path"
Description="Location of Freetype 2 library files"
Description="Specify whether to use the Hunspell spell checker library. This is included and requires no configuration."
/>
<BoolProperty
Name="AegisubUseFftw"
Category="Features"
DisplayName="Use FFTW"
Description="Specify whether to use the FFTW fourier transformation library to speed up audio spectrum rendering. You will need to configure the location of the FFTW library on the Library paths page."
Description="Specify whether to include support for the libass subtitle rendering library. You will need to configure the location of libass on the Library paths page."
Description="Specify whether to include support for CSRI subtitle rendering. The included helper library dynamically loads CSRI renderers, while linking another library allows you to e.g. link VSFilter.dll directly."
Category="Features"
>
<EnumValueName="HelperLibrary"DisplayName="Use included helper library"Description="Use the included helper library to dynamically load CSRI renderers."/>
<EnumValueName="External"DisplayName="Link other library"Description="Use another CSRI renderer"/>
<EnumValueName="false"DisplayName="No CSRI support"Description="Do not use CSRI"/>
</EnumProperty>
<StringProperty
Subtype="folder"
Name="CsriLibraryPath"
Category="Paths"
DisplayName="External CSRI library path"
Description="Location of external CSRI library files, for when not using the CSRI helper library."
Description="Make a special build of Aegisub, either enabling StartupLog to help end-users debug startup issues, or a Final Release whch displays version information differently."
Category="Features"
>
<EnumValueName="false"DisplayName="Make a regular build"Description="A regular developer build is made"/>
<EnumValueName="StartupLog"DisplayName="Start-up logging"Description="Show messages during start-up to figure out where an error occurs"/>
<EnumValueName="FinalRelease"DisplayName="Final release"Description="Shows version information in a manner suited for stable releases for general distribution."/>
</EnumProperty>
<StringProperty
Name="AegisubBuildCredit"
Category="BuildInfo"
DisplayName="Build credit"
Description="Name the build should be credited to. Appears in the about box, and in the title bar of development builds."
Description="Location of the SubWCRev.exe file used to process SVN working copies. Changes to this setting are stored in user settings, not in the main project file."
odrling