The existing hunk data structure does not contain any way to easily
store information about context surrounding the additions and removals
in a hunk. While this does work fine for normal diffs (where there is
never any surrounding context) this data structure is quite limiting for
other use cases.
Without support for surrounding context it is not possible to:
* Add support for unified or context format to the diff utility to
output surrounding context.
* Be able to implement a patch utility that uses the surrounding
context to reliably locate where to apply a patch when a hunk range
does not apply perfectly.
This patch changes Diff::Hunk such that its data structure more closely
resembles a unified diff. Each line in a hunk is now either a change,
removal, addition or context.
Allowing hunks to have context inside of them exposes that HackStudio
heavily relies on there being no context in the hunks that it uses for
its' git gutter implementation. The fix here is simple - ask git to
produce us a diff that has no context in it!
This avoids the overhead of allocating a new Array on every function
call, saving a substantial amount of time and avoiding GC thrash.
This patch only makes use of Op::Call in CallExpression. There are other
places we should codegen this op. We should also do the same for super
expression calls.
~5% speed-up on Kraken/stanford-crypto-ccm.js
Forcing every function call to allocate a new Array just to accommodate
spread parameters is not very nice, so let's start moving towards making
this a special case rather than the general (and only) case.
The var environments will unwind as needed with the ExecutionContext
and there's no need to include it in the unwind info.
We still need to do this for lexical environments though, since they
can have short local lifetimes inside a function.
This patch fixes an issue where the preview for an filter was always
based on the content bitmap even if the edit mode was set to Mask.
Now the preview is using the content or mask bitmap based on the
current edit mode and therefore generates the correct preview of
what is going to be changed.
This patch removes a todo where the revert for any changes could be
optimized. Previously every single pixel was copied back from the
reference bitmap to the content bitmap. Now the editors content
bitmap is just replaced with the reference bitmap that is a copy
of the unchanged content bitmap.
This patch adds a new Editing-Mask type to layers. This kind of mask
is used to restrict changes on the content bitmap only to areas where
the mask is drawn. The intensity of a pixel change is controlled by the
alpha-value of the mask.
Furthermore a function to invert and clear masks has been introduced.
When a new mask is created for a layer the edit mode of the layer is
also changed to Mask so that the user can immediately start to draw the
mask.
Performing these immediately can introduce a race condition between the
user's signal-related logic and LibLine's own, so defer the handlers to
make sure they run when our terminal IO cannot interfere with the
user's.
Since the relationship between VM and Bytecode::Interpreter is now
clear, we can have VM ask the Interpreter for roots in the GC marking
pass. This avoids having to register and unregister handles and
MarkedVectors over and over.
Since GeneratorObject can also own a RegisterWindow, we share the code
in a RegisterWindow::visit_edges() helper.
~4% speed-up on Kraken/stanford-crypto-ccm.js :^)
While the completion value of a variable declaration is specified to be
empty, we might already have a completion value in the accumulator from
a previous statement. Preserve it so as to avoid clobbering it.
This fixes 6 tests on test262.
This man page was referenced from some places. This is mostly a
condensed version of the POSIX behavior that the system call
implementation already has, only documenting the obviously visible
errors (in source code) we do actually report.
There are at most 576 granule samples/frequency lines, but the side data
can specify that the big_values granule type take up to 1024 samples.
The spec says in 2.4.3.4.6 that we should always stop reading Huffman
data once we have 576 samples, so that is what this change does. I also
add some useful spec comments while I'm here.
The SVGContext is a leftover from when SVG properties were more ad-hoc.
All properties are now (for better or worse) treated as CSS properties
(or handled elsewhere). This makes the SVGContext's fill/stroke
inheritance handling unnecessary.
This is a preparation before we can create a usable mechanism to use
filesystem-specific mount flags.
To keep some compatibility with userland code, LibC and LibCore mount
functions are kept being usable, but now instead of doing an "atomic"
syscall, they do multiple syscalls to perform the complete procedure of
mounting a filesystem.
The FileBackedFileSystem IntrusiveList in the VFS code is now changed to
be protected by a Mutex, because when we mount a new filesystem, we need
to check if a filesystem is already created for a given source_fd so we
do a scan for that OpenFileDescription in that list. If we fail to find
an already-created filesystem we create a new one and register it in the
list if we successfully mounted it. We use a Mutex because we might need
to initiate disk access during the filesystem creation, which will take
other mutexes in other parts of the kernel, therefore making it not
possible to take a spinlock while doing this.
This change was a long time in the making ever since we obtained sample
rate awareness in the system. Now, each client has its own sample rate,
accessible via new IPC APIs, and the device sample rate is only
accessible via the management interface. AudioServer takes care of
resampling client streams into the device sample rate. Therefore, the
main improvement introduced with this commit is full responsiveness to
sample rate changes; all open audio programs will continue to play at
correct speed with the audio resampled to the new device rate.
The immediate benefits are manifold:
- Gets rid of the legacy hardware sample rate IPC message in the
non-managing client
- Removes duplicate resampling and sample index rescaling code
everywhere
- Avoids potential sample index scaling bugs in SoundPlayer (which have
happened many times before) and fixes a sample index scaling bug in
aplay
- Removes several FIXMEs
- Reduces amount of sample copying in all applications (especially
Piano, where this is critical), improving performance
- Reduces number of resampling users, making future API changes (which
will need to happen for correct resampling to be implemented) easier
I also threw in a simple race condition fix for Piano's audio player
loop.
Assertion fails if the point is outside of the rect. This was introduced
in introduced in #18970 and causes serenity to crash when changing to 2x
resolution for a monitor, if the cursor after resizing is outside of
the new screen.
Added test to reproduce.
This turns all errors into either "OOM" or a proper shell error (if
propagation is impossible or meaningless).
Fixes `echo -en '\xfe\x4a' | $SHELL` crashing.
Before this change, we only blurred the shadows which would've been
generated without blur. This meant that we didn't generate a shadow when
the offset was zero, even with blur. However, other browsers generate a
shadow when blur is set.
Instead, always generate a rectangular contour of sufficient thickness,
blur it if needed and blit it into the right position at the very end.
Thanks to the blur radius, we'll have a shadow even when the offset is
zero.
This is a similar strategy to what v8 does. Use the ASAN API function
__asan_addr_is_in_fake_stack to check any fake stack frames associated
with each stack address we scan. This fully allows running test-js -g
with the option detect_stack_use_after_return turned on.
The valid range for temporal values (`nsMinInstant`/`nsMaxInstant`)
means performing nanosecond-valued integers could lead to an overflow.
NB: Only the `roundingMode: "day"` case was affected, as all others were
already performing the division on floating-point `fractional_second`
values. I'm adding `.0` suffixes everywhere to make this fact clearer.
This adds a few local tests as well, as those are tested with sanitizers
enabled by default, unlike test262.
`parse_a_calc_function_node()` only detects math functions, but
regularly encounters unrelated functions like `rgba()`. Since we now
support all the math functions, this message is just misleading spam.
In the merging of an extremely large navigables PR, a task queue source
was added after the unique task source start value that was supposed to
remain last in the enum.
Computing the table width algorithm bifurcates based on whether
table-root width is auto. We only adjust the used table width based on
cell percentage widths on the auto branch, thus the same check is needed
when we initialize cell widths.
This makes them trivially copyable, which is an assumption multiple
optimizations use when rebuilding the instruction stream.
This fixes most optimized crashes in the test262 suite.
See https://github.com/tc39/ecma262/commit/874ecf9
After this refactoring, we now correctly handle non-function /
non-undefined objects being passed multiple times: instead of skipping
assignment to promiseCapability altogether and failing with a
NotAFunction error in the end; on the second time the executor closure
is called, we return GetCapabilitiesExecutorCalledMultipleTimes.
This fixes the 7 `capability-executor-called-twice.js` test262 tests.
Cell percentage widths are relative to table width, not containing
block width. If the table width is auto, there isn't a normative
specification, only a brief mention that the user agent should try to
meet it.
As a starting point, we increase the width of the table such that it's
sufficient to cover min-width of cells with a percentage width. This
matches the behavior of other browsers, at least for simple cases.
We do this by moving the `LoadImmediate undefined` instruction to a
separate basic block which jumps to the case's block unconditionally.
We enter a case initially using this wrapper, but when falling through,
we directly jump to the next case's block.
This adds support for exporting class expressions, which was previously
TODO'd.
We now correctly set the binding name of exports to `"*default*"` if
they are unnamed. I'm not sure what the difference between the
`InitializationMode` kinds is, but using `Initialize` fixes a bunch of
tests.
Note that some export tests (e.g. `eval-export-dflt-expr-cls-named.js`)
still fail, as we don't set the "name" property of exported classes
correctly.
176 new passes on test262
The immutability of the string is not relevant here, since the string
we're given was allocated in the IPC serialization layer and will be
destroyed shortly afterwards. Additionally, noone relies on
DeprecatedString-specific functionality. This will make it easier to
convert the IPC layer itself to String later on.
This removes a lot of duplicated stream creation code from the plugins,
and also simplifies the way that the appropriate plugin is found. This
mirrors the ImageDecoderPlugin design and necessitates new sniffing
methods on the loaders.
While this would be useful in the future for implementing a multi-tiered
optimization strategy, currently a binary on/off is enough for us. This
removes the confusingly on-by-default `OptimizationLevel::None` option
which made the optimization pipeline a no-op even if
`Bytecode::Interpreter::set_optimizations_enabled` had been called.
Fixes#15982
Since https://reviews.llvm.org/D131441, libc++ must be included before
LibC. As clang includes libc++ as one of the system includes, LibC
must be included after those, and the only correct way to do that is
to install LibC's headers into the sysroot.
Targets that don't link with LibC yet require its headers for one
reason or another must add install_libc_headers as a dependency to
ensure that the correct headers have been (re)installed into the
sysroot.
LibC/stddef.h has been dropped since the built-in stddef.h receives
a higher include priority.
In addition, string.h and wchar.h must
define __CORRECT_ISO_CPP_STRING_H_PROTO and
_LIBCPP_WCHAR_H_HAS_CONST_OVERLOADS respectively in order to tell
libc++ to not try to define methods implemented by LibC.
Once LibC is installed to the sysroot and its conflicts with libc++
are resolved, including LibC headers in such a way will cause errors
with a modern LLVM-based toolchain.
This is needed to avoid including LibC headers in Lagom builds.
Unfortunately, we cannot rely on the build machine to provide a
fully POSIX-compatible ELF header for Lagom builds, so we have to
use our own.
libc++ disallows including LibC's complex.h in C++ mode. This means that
a C++ file cannot expect LibC's complex.h to be included, and thus
cannot use c-prefixed complex number functions. As a result,
complex.cpp is broken when libc++ has a higher include priority
than LibC.
A check for __cplusplus has been added to complex.h to warn users of
toolchains that don't use libc++.
If an exception is thrown by FunctionDeclarationInstantiation for an
async or async-generator function, we still need to return a promise.
We can't just throw the exception.
81 new passes on test262. :^)
Previously, strings would exit immediately if there was an error
changing file ownership. We now print an error to stderr and
continue when an error occurs.
Previously, the `-p` option printed the path of the file being
processed before any strings for that file. The `-f` prints the file
path before each string . This matches the behavior of strings on
Linux and FreeBSD.
Currently the only error that can happen is an OOM. However, in the
future there may be other errors that this function may throw, such as
detecting an invalid patch.
This ensures that min-content contributions from cells with no content
are computed using their calculated values, which are never considered
for min-content before then. The specification diverges from column
measures algorithm, which doesn't use specified width of cells anywhere.
The exponent might be larger than the range of values representable by
an i32, so we have to use the `fmod` function instead of the modulo
operator.
This fixes 3 test262 tests on AArch64. No changes on x86-64.
The observer callbacks can do all kinds of things, so let's not be in
the middle of iterating the set in case someone decides to mutate it.
Fixes a crash when loading https://lichess.org/
This prototype is a bit tricky in that we need to maintain the iteration
state of the mapped iterator's inner iterator as we return values to the
caller. To do this, we create a FlatMapIterator helper to perform the
steps that apply to the current iteration state.
This uses a new Iterator type called IteratorHelper. This does not
implement IteratorHelper.prototype.return as that relies on generator
objects (i.e. the internal slots of JS::GeneratorObject), which are not
hooked up here.
Iterator.from creates an Iterator from either an existing iterator or
an iterator-like object. In the latter case, it sets the prototype of
the returned iterator to WrapForValidIteratorPrototype to wrap around
the iterator-like object's iteration methods.
This is in preparation for an upcoming implementation of the Iterator
Helpers proposal. That proposal will require a JS::Object type named
"Iterator", so this rename is to avoid conflicts.
`seek_demuxer_to_most_recent_keyframe()` wasn't correctly returning in
cases where an error was thrown by the demuxer. To avoid this, the
function now returns the error, and the playback state handler must act
on it instead, allowing it to exit the seeking state early.
The EBML specification allows for CRC32 elements to be placed as the
first child element of a master element. However, our parsing of master
elements didn't take that into account, so an error would be thrown.
Instead of erroring out, the `parse_master_element()` function will now
skip CRC32 elements that are found as the first child of a master
element. If it is found after the first child, that will be considered
an error.
Void elements will also be skipped by `parse_master_element()`.
Since the `parse_cluster()` function has to seek the stream back to the
cluster's first child in order to allow cues' positions to be used
correctly, `parse_master_element()` had to be changed to return the
first element position, since the callback is not invoked for CRC32
elements. This means that the parameter used to communicate the element
position to the child element parsing function is unused, so that is
removed.
The fact that side effects clobbers the accumulator isn't just annoying
inside the instruction handler, but also really counter-intuitive in
the bytecode program itself.
17 new passes on test262. :^)
Instead of assuming that there's an active AST interpreter, this code
now takes VM& everywhere and invokes the appropriate interpreter.
92 new passes on test262. :^)
The CSS box-shadow property takes 2-4 properties that are `<length>`s,
those being:
- offset-x
- offset-y
- blur-radius
- spread-radius
Previously these were resolved directly to concrete Lengths at parse
time, but now they will be parsed as LengthStyleValues and/or
CalculatedStyleValues and be stored that way until styles are later
resolved.
The used width is already a content width, which doesn't include
borders. Border widths should be subtracted from the specified width
instead, since that initially specifies the total width including
borders, for consistent comparison. Also handle table box padding as an
additional fix.
This solution is a middle ground between re-computing `cos` every time
and a much more mathematically complicated approach (as we have in the
decoder).
While still being far from optimal it already gives us a 10x
improvement, not that bad :^)
Co-authored-by: Tim Flynn <trflynn89@pm.me>
This is a sensible separation of concerns that mirrors the WindowServer
IPC split. On the one hand, there is the "normal" audio interface, used
for clients that play audio, which is the primary service of
AudioServer. On the other hand, there is the management interface,
which, like the WindowManager endpoint, provides higher-level control
over clients and the server itself.
The reasoning for this split are manifold, as mentioned we are mirroring
the WindowServer split. Another indication to the sensibility of the
split is that no single audio client used the APIs of both interfaces.
Also, useless audio queues are no longer created for managing clients
(since those don't even exist, just like there's no window backing
bitmap for window managing clients), eliminating any bugs that may occur
there as they have in the past.
Implementation-wise, we just move all the APIs and implementations from
the old AudioServer into the AudioManagerServer (and respective clients,
of course). There is one point of duplication, namely the hardware
sample rate. This will be fixed in combination with per-client sample
rate, eliminating client-side resampling and the related update bugs.
For now, we keep one legacy API to simplify the transition.
The new AudioManagerServer also gains a hardware sample rate change
callback to have exact symmetry on the main server parameters (getter,
setter, and callback).
This facility was added in 15a1d9a, but isn't being used for anything.
It wasn't even hooked up to LibGUI for applications to use.
Relevant use-cases, such as the most prominent one in `AnalogClock`, use
`GUI::Window::set_frameless()` instead.
From what I can tell, this facility was added to WSWindow/GWindow in
2019 in 9b71307. I only found a single place in the codebase still using
this facility: `WindowServer::Menu::start_activation_animation()`. A
subtle fade-out animation that happens when a menu item is selected, and
the menu disappears.
I think our compositing facilities have improved enough to make this
facility redundant. The remaining use mentioned above was ported to just
directly blit the fade-out animation instead of requesting it from
WindowServer.
While this is a useful piece of information it means that diff is
producing hunks that are not of a valid normal diff format. This breaks
the ability to redirect the output of diff to a file to generate a
patch.
Previously, an audio loader could succeed for an HTMLVideoElement that
contains a video file, which caused the duration to be set to the bogus
duration of the audio loader instead of the correct duration from the
video container. Instead of setting the duration based on audio always,
set it to the video duration if we are creating a video element.
On style update, we have to preserve the invariant established when we
built the layout tree - some properties are applied to the table wrapper
and the table box values are reset to their initial values.
This also ensures that the containing block of a table box is always a
table wrapper, which isn't the case if we set absolute position on the
box instead of the wrapper.
Fixes#19452.
This is only needed because currently spec doesn't explicitly define
that navigable should have a pointer to associated container and when
this pointer should be set.