While macOS backtrace(3) puts a space directly after the mangled symbol
name, some versions of glibc put a + directly after it. This new logic
accounts for both situations when trying to demangle.
Co-Authored-By: Andreas Kling <kling@serenityos.org>
This ports an implementation of the FTS functions that can be used to
traverse the file system. They are non-standard, but provided by glibc
and most BSD systems. This ported library implements FTS for musl-based
Linux systems and happens to work on Serenity.
We would start the timer to send playback time updates to the element
before playback had started, so in cases where the `HTMLMediaElement`
(incorrectly) creates both an `AudioTrack` and `VideoTrack`, it will
not cause the seek bar to flicker between the current video position
and zero.
Which for now will just call the DeprecatedString version of this
function. This is intended to be used in porting code over to using the
new String equivalent with the end goal of removing the DeprecatedString
version of this function.
This allows us to port a whole heap of IDL interfaces from
DeprecatedString to String.
NewAKString is effectively the default for any new IDL interface, so
let's mark this as the default behavior. It also makes it much easier to
figure out whatever interfaces are still left to port over to new AK
String.
Currently, they are not extremely useful, but the plan is to store
all function-local state in JSSpecCompiler::Function and all
"translation unit" state in ExecutionContext.
Currently, every public DOM::Element method which changes an attribute
fires this handler itself. This was missed in commit 720f7ba, so any
user of that API would not fire the internal handler.
To fix this, and prevent any missing invocations in the future, invoke
the handler from from Attr::handle_attribute_changes. This method is
reached for all attribute changes, including adding/removing attributes.
This ensures the handler will always be fired, and reduces the footprint
of this ad-hoc behavior.
Note that our ad-hoc handler is not the "attribute change steps" noted
by the spec. Those are overridden only by a couple of specific elements,
e.g. HTMLSlotElement. However, we could easily make our ad-hoc handler
hook into those steps in the future.
Instead of trying to keep a live reference to the bytecode interpreter's
current instruction stream iterator, we now simply copy the current
iterator whenever pushing to the ExecutionContext stack.
This fixes a stack-use-after-return issue reported by ASAN.
The arguments are passed on registers, so if we pass only 3 defined
arguments then the fourth argument for the prctl syscall could have
garbage value within it.
To avoid possible bugs, always pass 3 arguments to a raw syscall prctl
call in addition to the prctl sub-option (the first argument).
Before, when looking in /bin in File Manager, the About application's
icon was missing. Now it is a serenity_app instead of serenity_bin the
icon is visible!
This works by adding source start/end offset to every bytecode
instruction. In the future we can make this more efficient by keeping
a map of bytecode ranges to source ranges in the Executable instead,
but let's just get traces working first.
Co-Authored-By: Andrew Kaster <akaster@serenityos.org>
These return true if the last modification time, last access time or
creation time of a file is greater than the given reference file.
If the `-L` option is in use and the given reference file is a
symbolic link then the timestamp of the file pointed to by the
symbolic link will be used.
The matrix used in the spec is column-major but Gfx::Matrix4x4 is
row-major so we need to transpose the values. This will fix internal
operations on that matrix. Because we also transposed the readonly
matrix property getters the matrix is again transposed when reading
so the JavaScript world only sees a column-major matrix.
MAP_FILE is not in POSIX, and is simply in most LibCs as a "default"
mode. Our own LibC defines it as 0, meaning "no flags". It is also not
defined in some OS's, such as Haiku. Let's be more portable and not use
the unnecessary flag.
Before this change, we were creating a new anonymous flex item for every
inline-level child of a flex container, even when we had a sequence of
inline-level children.
The fix here is to simply keep putting things in the last child of the
flex container, if that child is already an anonymous flex item.
This is a bit iffy, but since <br> elements can't be implemented in
"just CSS" today, we should also exclude them from the blockification
algorithm. This is important, since <br> is expected to always have
inline-like behavior.
Let users save their nix develop derivation in a profile by ignoring
this specific folder. It encourages the following workflow:
```
nix develop Toolchain/ --profile Toolchain/nix-profiles/dev
```
Which stops the dev enviornment being collected in the nix store. Later
devs can come back and do:
```
nix develop Toolchain/nix-profiles/dev
```
To continue where they left off, without having to download everything
from nixpkgs again.
xlibswrapper is removed from `Toolchain/serenity.nix` as it is a
wrapper package and has been deprecated. We don't use it in the build
anyway.
This fixes#19286
During the build, python3 is required so we add `python3` as a build
dependency.
Add a nix flake to `Toolchain/` that wraps the existing nix derivation
`Toolchain/serenity.nix`. This also comes with a lockfile making the nix
developer enviornment setup more reproducible.
We also update the documentation for "other builds" to mention the
flake.
On platforms that support it, enable using ``<execinfo.h>`` to get
backtrace(3) to dump a backtrace on assertion failure. This should make
debugging things like WebContent crashes in Lagom much easier.
There is only one, width/height -> aspect-ratio. This brings us
very slightly closer to spec and triggers a re-layout after
updating these values from JavaScript, which wasn't the case
before.
If stop_timer() is called from ~EventReceiver(), then the virtual
functions will end up calling the overload from the base class. As
EventReceiver::class_name() is pure virtual, this calls
__cxa_pure_virtual and crashes. We should not be calling virtual
functions from the destructor, and especially not pure virtual ones.
This "fixes" a crash in Piano, but the root cause of the problem is
still unfixed.
d06d4eb made the `clip` property apply to children of an absolute-
positioned element, but caused it not to be applied to the element the
property was applied to directly.
To fix this, apply the clip in new `before_paint()` and `after_paint()`
functions. Doing so keeps painter state from leaking from `paint()`,
but still allows subclasses of `PaintableBox` clip their contents
correctly without repeating the application of the clip rectangle.
