It's not safe to allocate from the GC heap while in the constructor of a
GC heap cell. (Because if this ends up triggering a collection, we may
end up trying to call through an uninitialized vtable).
This was already done safely in the initialize() virtual in much of
LibJS and LibWeb. This patch moves the logic for prototypes, mixins,
and CSSStyleDeclaration as well.
Fixes a long-standing GC crash that was pretty easy to reproduce by
refreshing https://vercel.com/
This completion only works if you have lagom already built in some
capacity, since it scans the build directory tree for binaries, removing
known false positives. However, that is both more accurate than asking
ninja for the targets and filtering those, and it also makes it
independent of the build system used.
This is very convenient for anyone like me who regularly runs the Clang
toolchain. The toolchain is not completed for Lagom and the
toolchain-independent help command.
This in turn enables `./Meta/serenity.sh test aarch64` and the CI
scripts to work with the AArch64 port.
As the RPi doesn't have a debugcon-like device, we create two serial
devices. The system console, UART0 is redirected to `debug.log`, while
UART1 is made available to the userspace and is used as the stdout for
the test runner script.
We are not yet able to run the full test suite, as the kernel panics due
to some unimplemented features.
Note that Qemu `master` or our patched Qemu build is required for
`SystemServer` to recognize the `system_mode=self-test` parameter.
Newer versions of QEMU prevent the user from running a GL-rendered
display while a SPICE display is active due to incompatibilities.
Since there is no way to disable QEMUs (apparently implicit) SPICE
display, make sure that we only enable SPICE support if the user
requested running with SPICE specifically. In this case, QEMU picks the
default SPICE client instead of rendering a display using whatever our
default on that platform would be.
We already install C++ source files to allow debugging applications
in HackStudio.
Installing GML files can make editing application widgets a bit faster
and easier, as you no longer need to copy files to the system. :^)
This little program allows us to take the NetworkSettings app away
from being an elevated GUI app.
It receives a JsonObject on STDIN and writes it to the global
Network configuration file.
If the write was successfull it will apply the changes.
We now load SVG icons (via the Qt resource system) and render them into
a QIcon (with normal and disabled variants) using system colors.
We also re-render them if the system color theme changes.
This instantly makes Ladybird look less foreign on my Linux box.
I drew the icons myself, and they could definitely be more optimized,
but this was my first time using Inkscape. :^)
GCC 13 was released on 2023-04-26. This commit fixes Lagom build errors
when using an updated host toolchain:
- Adds a workaround for a bug in constraint handling, which made LibJS
fail to compile: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109683
- Silences the new `-Wdangling-reference` diagnostic globally. It
produces multiple false positives with no clear way to silence them
without `#pragmas`.
- Silences `-Wself-move` in `RefPtr` tests as GCC 13 adds this
previously Clang-exclusive warning.
This ensures that all visit_edges implementations include a call to
Base::visit_edges. In particular, this gives three nice benefits:
- The call can't be forgotten (the main benefit, of course).
- All of the calls look the same. In other words, always use "Base"
instead of the actual concrete class.
- Ensure the object has a call to JS_CELL or JS_OBJECT in the
definition. Otherwise, Base will not be defined and the call will
not compile.
Some hardware/software configurations crash KVM as soon as we try to
start Serenity. The exact cause is currently unknown, so just fully
revert it for now.
This reverts commit 897c4e5145.
The new baked image is a Prekernel and a Kernel baked together now, so
essentially we no longer need to pass the Prekernel as -kernel and the
actual kernel image as -initrd to QEMU, leaving the option to pass an
actual initrd or initramfs module later on with multiboot.
That pattern seems to show up a lot in code written by people that
aren't intimately familiar with the lifetime model of Error and Strings.
This commit makes the compiler detect it and present a more helpful
diagnostic than "garbage string at runtime".
This class can be used to run a task in another thread, and allows the
caller to wait for the task to complete to retrieve any error that may
have occurred.
Currently, it doesn't support functions returning a value on success,
but with some template magic that should be possible. :^)
For the most part no behavior change, except that we now pass
-Wno-implicit-const-int-float-conversion and -Wno-literal-suffix
only to clang and gcc each in both lagom and serenity builds,
while we previously passed them to both in lagom builds (and
passed them to one each in serenity builds). The former is
a clang flag, the latter a gcc flag, but since we also use
-Wno-unknown-warning-option it doesn't really matter.
In CLDR 42 and earlier, we were able to assume all cldr-localename files
existed for every locale. They now do not exist for locales that don't
provide any localized data. Namely, this is the "und" locale (which is
an alias for the root locale, i.e. the locale we fall back to when a
user provides an unknown locale).
Further, we were previously able to assume that each currencies.json in
cldr-numbers contained all currencies. This file now excludes currencies
whose localized names are the same as the currency key. Therefore, we
now preprocess currencies.json to discover all currencies ahead of time,
much like we already do for languages.json.
Some of these are allocated upon initialization of the intrinsics, and
some lazily, but in neither case the getters actually return a nullptr.
This saves us a whole bunch of pointer dereferences (as NonnullGCPtr has
an `operator T&()`), and also has the interesting side effect of forcing
us to explicitly use the FunctionObject& overload of call(), as passing
a NonnullGCPtr is ambigous - it could implicitly be turned into a Value
_or_ a FunctionObject& (so we have to dereference manually).
Previously, the condition was reversed, so we would stop immediately on
a file that has at least one working chunk, and we would infinitely loop
on a file with no chunks.
Target GDB is only used for debugging the kernel, which is not relevant
to most people. Starting with 924758c6f8, GDB would always be built
as part of the toolchain if the user didn't have it installed. This is
unnecessary.
This commit adds a separate script for building GDB, which the user
needs to explicitly invoke. A message is added to Meta/debug-kernel.sh
which alerts the user to this fact.
When we had 32 bit support in the OS kernel and userland, the very bare
minimum CPU we supported was Pentium 3, but now the CPU is just required
to support x86-64 long mode to be supported, so the exact model is not
very important.
I chose the QEMU64 virtual CPU model, because the whole concept of the
QEMU ISA-PC machine is that it checks how the kernel handles arbitrarily
old hardware setup.
Loads of changes that are tightly connected... :/
* Change lambdas to static functions
* Add spec docs to those functions
* Keep the current scope around as a parameter
* Add wrapping classes for some Certificate members
* Parse ec and ecdsa data from certificates
Previously, we would unconditionally build GDB from source for the
AArch64 toolchain. This commit makes it possible to use the system's
`gdb` binary if it supports the architecture, or `aarch64-elf-gdb` if
such a package is installed.
An `aarch64-elf-gdb` package will be available through Homebrew once
this PR is merged: https://github.com/Homebrew/homebrew-core/pull/127323
This solves the longhands must be initialized before shorthands
problem by just initializing the value when it is first requested.
This will be needed for background-position-x/y which is a longhand
of a longhand, which otherwise breaks the longhands before shorthands
rule.
This makes it easy to compare the performance of Serenity's
deflate implementation to the host system implementation.
On my M1 Max MBP:
% time gunzip -c \
/Users/thakis/Downloads/trace_bug.json.gz > /dev/null
takes between 0.064s and 0.082s.
% time Build/lagom/gunzip -c \
/Users/thakis/Downloads/trace_bug.json.gz > /dev/null
on the other hand takes 4 seconds.
(That .gz file is a 5.8M file -- the largest .gz file that happend to be
in my Downloads folder.)
This is an implementation that tries to follow the spec as closely as
possible, and works with Qemu's Intel HDA and some bare metal HDA
controllers out there. Compiling with `INTEL_HDA_DEBUG=on` will provide
a lot of detailed information that could help us getting this to work
on more bare metal controllers as well :^)
Output format is limited to `i16` samples for now.
`SERENITYOS` is also set when compiling Lagom on SerenityOS, so we can't
just check it and expect `CMAKE_STAGING_PREFIX` to be set.
Instead, check `CMAKE_STAGING_PREFIX` directly and use that as an
indicator for whether we can install a file there.
It appears that QEMU on macOS doesn't have the VirtIO GPU variants that
support VGA functionality. Those variants are not especially important
to us, because we don't use any kind of VGA functionality in our kernel
anyway.
Therefore, for macOS, we could decide to use virtio-gpu-gl-pci and
virtio-gpu-pci devices instead.
Required by code that brand checks native constructors.
For example, Wistia brand checks XMLHttpRequest by doing:
```
XMLHttpRequest.prototype.constructor.toString()
```
It then checks if it matches either one of:
```
function XMLHttpRequest() { [native code] }
```
```
[object XMLHttpRequestConstructor]
```
If neither matches, it disables HLS playback and prints:
"The XMLHttpRequest constructor has been tampered with. Because this
affects CORS/Range XHR requests, HLS playback has been disabled.
To enable HLS playback and other important features, please remove code
that changes the definition of window.XMLHttpRequest."
We hit this path due to not giving generated constructors a name, as
we would provide `function () { [native code] }`.
This patch parses enough of GPOS tables to be able to support the
kerning information embedded in Inter.
Since that specific font only applies positioning offsets to the first
glyph in each pair, I was able to get away with not changing our API.
Once we start adding support for more sophisticated positioning, we'll
need to be able to communicate more than a simple "kerning offset" to
the clients of this code.
Currently, the WebAssemblyObject implements a visitor to keep its static
objects alive. This custom attribute will be used to hook the generated
namespace object's visitor to one that we define in non-generated code.
Interfaces with a LegacyNamespace extended attribute should have their
constructors defined on the namespace identified by the LegacyNamespace
attribute value.
This is used by WebAssembly IDL files. For now, we mostly use this for
error messages and cache keys (to ensure compatibility with existing
code as WebAssembly is ported to IDL).
For example, consider the attribute:
interface Element {
[PutForwards=value] readonly attribute DOMTokenList classList;
}
When `classList` is set, we should instead set the attribute `value` on
the `classList` attribute of the Element interface.
Don't try to reserve capacity for a variadic arguments list unless we
actually have enough arguments to fill it with anything. Otherwise we
may overflow to an extremely large size if, e.g., the argument count
is 0 and the start of the variadic arguments is index 1.
We can always read the basic format information (sample rate, bit depth,
etc.), but we will also print artist, album, and title if available in
the metadata.
Similar to POSIX read, the basic read and write functions of AK::Stream
do not have a lower limit of how much data they read or write (apart
from "none at all").
Rename the functions to "read some [data]" and "write some [data]" (with
"data" being omitted, since everything here is reading and writing data)
to make them sufficiently distinct from the functions that ensure to
use the entire buffer (which should be the go-to function for most
usages).
No functional changes, just a lot of new FIXMEs.
Before, some loader plugins implemented their own buffering (FLAC&MP3),
some didn't require any (WAV), and some didn't buffer at all (QOA). This
meant that in practice, while you could load arbitrary amounts of
samples from some loader plugins, you couldn't do that with some others.
Also, it was ill-defined how many samples you would actually get back
from a get_more_samples call.
This commit fixes that by introducing a layer of abstraction between the
loader and its plugins (because that's the whole point of having the
extra class!). The plugins now only implement a load_chunks() function,
which is much simpler to implement and allows plugins to play fast and
loose with what they actually return. Basically, they can return many
chunks of samples, where one chunk is simply a convenient block of
samples to load. In fact, some loaders such as FLAC and QOA have
separate internal functions for loading exactly one chunk. The loaders
*should* load as many chunks as necessary for the sample count to be
reached or surpassed (the latter simplifies loading loops in the
implementations, since you don't need to know how large your next chunk
is going to be; a problem for e.g. FLAC). If a plugin has no problems
returning data of arbitrary size (currently WAV), it can return a single
chunk that exactly (or roughly) matches the requested sample count. If a
plugin is at the stream end, it can also return less samples than was
requested! The loader can handle all of these cases and may call into
load_chunk multiple times. If the plugin returns an empty chunk list (or
only empty chunks; again, they can play fast and loose), the loader
takes that as a stream end signal. Otherwise, the loader will always
return exactly as many samples as the user requested. Buffering is
handled by the loader, allowing any underlying plugin to deal with any
weird sample count requirement the user throws at it (looking at you,
SoundPlayer!).
This (not accidentally!) makes QOA work in SoundPlayer.
headless-browser currently uses its own PageClient to load web pages
in-process. Due to this, it also needs to set up a whole bunch of other
objects needed to run LibWeb, e.g. image decoders, request servers, etc.
This changes headless-browser to instead implement a WebView to launch
WebContent out-of-process. This implementation is almost entirely empty,
but can be filled in as-needed. For example, we may want to print
JavaScript console messages.
At the moment, all it can do is read all image formats that LibGfx can
read and save to any image format that LibGfx can write (currently bmp,
png, qoi).
Currently, it drops all image metadata (including color profiles).
Over time, this could learn tricks like keeping color profiles,
converting an image to a different color profile, cropping out a part of
an image, and so on.
"image" was an alias for "qemu-image".
I want to add an `image` userland utility, which clashes with that
shortname.
So remove the existing "image" target. It was just an alias for
"qemu-image".
If you use serenity.sh to build, nothing changes. This only affects you
if you run ninja manually -- you now have to say `ninja qemu-image` to
build the disk image.
Commit: 2c84466ad8 ("Kernel/Storage: Introduce new boot device
addressing modes") changed the way we pass the boot device parameter.
That commit missed updating boot parameter in the run.sh script for NVMe
boot devices.
We should be able to run this locally, as long as ENABLE_LAGOM_LADYBIRD
is true, or if building ladybird from the ladybird source directory.
This removes a special case from the Lagom CI yml file.
Add some prose to the introduction of Lagom about how we use it. Also,
move the section on including Lagom in other projects above the fuzzer
documentation.
Remove the explicit cmake commands from the Fuzzer documentation, as the
script should be the source of truth.
When an IDL file has #imports and the IDL interface exposes an iterator,
the bindings generator would generate #include statements missing the
class name of the iterator in the form 'LibWeb/{namespace}/Iterator'.
This change only generates the iterator #include statement for the top
interface that is the iterator.
These are treated differently as the interface members are placed on the
object itself, not its prototype.
As the object itself still needs to be hand-written code, and we can no
longer fully hide the gnarly generated code in the prototype object,
these now generate a 'mixin' class that is added to the actual object
through inheritance.
https://webidl.spec.whatwg.org/#Global
The Window object is massive, so let's do the conversion to IDL step
by step. First up: getting rid of the manual constructor and prototype
definitions, which can be generated from an empty `interface Window`.
This relied on pulling the current realm from the main thread VM, which
requires an execution context to be on the VM's stack. This heavily
relied on the dummy execution context that is always on the stack, for
example, when parsing the UA style sheets where no JavaScript is
running.
This class had slightly confusing semantics and the added weirdness
doesn't seem worth it just so we can say "." instead of "->" when
iterating over a vector of NNRPs.
This patch replaces NonnullRefPtrVector<T> with Vector<NNRP<T>>.
This is implemented as a Clang frontend tool, and currently does two
things:
- Ensure for all fields wrapped in {Nonnull,}GCPtr<T>, T inherits from
JS::Cell
- Ensure for all fields not wrapped in {Nonnull,}GCPtr, that the type
does not inherit from JS::Cell (otherwise it should be wrapped in a
Ptr class).
In the future, this tool could be extended further. For example, we may
consider validating all implementations of Cell::visit_impl.
This also removes DirIterator::error_string(), since the same strerror()
string will be included when you print the Error itself. Except in `ls`
which is still using fprintf() for now.
This sorts the array of generated emoji data by code point (first by
code point length, then by code point value). This lets us use a binary
search to find emoji data, rather than the current linear search.
In a profile of scrolling around /home/anon/Documents/emoji.txt, this
reduces the runtime of Gfx::Emoji::emoji_for_code_points from 69.03% to
28.42%. Within that, Unicode::find_emoji_for_code_points reduces from
28.42% to just 1.95%.
These symlinks' only purpose was to be copied into the rootfs along with
the rest of Base. Instead of storing symlinks to files that either
don't exist in the Base directory, or point to an absolute path outside
of the serenity folder, move these symlinks into the
build-root-filesystem.sh script.
Similar to the FontDatabase, this will be needed for Ladybird to find
emoji images. We now generate just the file name of emoji image in
LibUnicode, and look for that file in the specified path (defaulting to
/res/emoji) at runtime.
`consume_until(foo)` stops before foo, and so does
`ignore_until(Predicate)`, so let's make the other `ignore_until()`
overloads consistent with that so they're less confusing.
This commit moves the implementation of getopt into AK, and converts its
API to understand and use StringView instead of char*.
Everything else is caught in the crossfire of making
Option::accept_value() take a StringView instead of a char const*.
With this, we must now pass a Span<StringView> to ArgsParser::parse(),
applications using LibMain are unaffected, but anything not using that
or taking its own argc/argv has to construct a Vector<StringView> for
this method.
10ms (the default) is ridiculous and causes all kinds of glitches if we
actually want to have a low-latency queue.
<https://gitlab.com/qemu-project/qemu/-/issues/1076#note_996636777>
suggests 2ms (and no lower than 1ms). This improves audio glitch
resistance at our current 512 sample buffer size, but going lower is
still not possible.
Imported functions in Wasm may throw JS exceptions, and we need to
preserve these exceptions so we can pass them to the calling JS code.
This also adds a `assert_wasm_result()` API to Result for cases where
only Wasm traps or values are expected (e.g. internal uses) to avoid
making LibWasm (pointlessly) handle JS exceptions that will never show
up in reality.
This adds the condition member.type->is_string() to the if statement, so
that we now conditionally check the dictionary member is a new string
and associated with an optional constructor parameter.
The LibWeb fuzzer build is really slow, so for local builds it is useful
to disable it when you're not interested in running that fuzzer.
Co-authored-by: Andrew Kaster <akaster@serenityos.org>
At the moment, this processes the RIFF chunk structure and extracts
the ICCP chunk, so that `icc` can now print ICC profiles embedded
in webp files. (And are image files really more than containers
of icc profiles?)
It doesn't even decode image dimensions yet.
The lossy format is a VP8 video frame. Once we get to that, we
might want to move all the image decoders into a new LibImageDecoders
that depends on both LibGfx and LibVideo. (Other newer image formats
like heic and av1f also use video frames for image data.)
This picks up any upstream brew changes that haven't been pulled into
the default Azure VMs. For example, the ffmpeg dependency of qt is
currently broken in the 8 day old brew commit they are using.
Make sure that we set SERENITY_SOURCE_DIR in ctest, and make sure to
pass the test root to the CI job.
More overhaul of test-js 'test root' finding is needed however.
When a constructor has an optional dictionary as argument, and those
members are of type new string, make sure that we release_value()
setting the dictionary members.
The patch also contains modifications on several classes, functions or
files that are related to the `JPGLoader`.
Renaming include:
- JPGLoader{.h, .cpp}
- JPGImageDecoderPlugin
- JPGLoadingContext
- JPG_DEBUG
- decode_jpg
- FuzzJPGLoader.cpp
- Few string literals or texts
This makes use of the new [UseNewAKString] extended attribute. Using
Vector storage will make it easier to make this interface into an IDL
iterable. It seems the reason it didn't use Vector originally was due
to awkward DeprecatedString -> String conversions.
Adding the [UseNewAKString] extended attribute to an interface will
cause all IDL string types to use String instead of DeprecatedString.
This is done on an per interface level instead of per type/parameter
because:
- It's much simpler to implement, as the generators can already access
the interface's extended attributes. Doing it per type/parameter
would mean parsing and piping extended attributes for each type that
doesn't already take extended attributes, such as unions.
- Allows more incremental adoption of AK::String. For example, adding
[UseNewAKString] to BodyInit would require refactoring Request,
Response and XMLHttpRequest to AK::String in one swoop. Doing it on
the interface allows you to convert just XHR and its dependencies at
once, for example.
- Simple string return types (i.e. not parameterised or not in a union)
already accept any of the string types JS::PrimitiveString::create
accepts. For example, you can add [UseNewAKString] to DOMStringMap to
convert Element attributes to AK::String and still return AK::String
from get_attribute, without adding [UseNewAKString] to Element.
- Adding [UseNewAKString] to one function typically means adding it to
a bunch of other functions, if not the rest of them. For example,
adding [UseNewAKString] to the parameters FormData.append would
either mean converting AK::String to AK::DeprecatedString or storing
the AK::String as-is, making the other functions of FormData have to
convert back from AK::String or also support AK::String.
This includes an Error::create overload to create an Error from a UTF-8
StringView. If creating a String from that view fails, the factory will
return an OOM InternalError instead. VM::throw_completion can also make
use of this overload via its perfect forwarding.
The aarch64 processor is set up to trap on unaligned memory accesses, so
to enforce that the compiler correctly generates aligned accesses, the
-mstrict-align flag is needed. We also need the -Wno-cast-align as there
are some files in AK that don't build without the flag.
The parser is still very much a work-in-progress, but it can currently
parse most of the basic bits, the only *completely* unimplemented things
in the parser are:
- heredocs (io_here)
- alias expansion
- arithmetic expansion
There are a whole suite of bugs, and syntax highlighting is unreliable
at best.
For now, this is not attached anywhere, a future commit will enable it
for /bin/sh or a `Shell --posix` invocation.
For example, consider cases where we want to propagate errors only in
specific instances:
auto result = read_data(); // something like ErrorOr<ByteBuffer>
if (result.is_error() && result.error().code() != EINTR)
continue;
auto bytes = TRY(result);
The TRY invocation will currently copy the byte buffer when the
expression (in this case, just a local variable) is stored into
_temporary_result.
This patch binds the expression to a reference to prevent such copies.
In less trival invocations (such as TRY(some_function()), this will
incur only temporary lifetime extensions, i.e. no functional change.
First, this adds an overload of PrimitiveString::create for StringView.
This overload will throw an OOM completion if creating a String fails.
This is not only a bit more convenient, but it also ensures at compile
time that all PrimitiveString::create(string_view) invocations will be
handled as String and OOM-aware.
Next, this wraps all invocations to PrimitiveString::create(string_view)
with MUST_OR_THROW_OOM.
A small PrimitiveString::create(DeprecatedFlyString) overload also had
to be added to disambiguate between the StringView and DeprecatedString
overloads.
This reverts commit b0606d90f0.
This seems to prevent libegl-mesa0 from being installed (which for some
reason isn't failing the Azure jobs - the failure seen later is that
ccache is not installed).
Previously the tool was removing the Root directory after configuring
cmake which was breaking the build, as some cmake rules put some
necessary files there.
Moving the cmake command to be the last one makes it regenerate those
files automatically. :^)
Setting the DWARF version after having selected which level of debug
information to generate apparently undoes some settings again.
Doing the reverse apparently keeps both the version and the debug level
setting, resulting in a significantly smaller disk image size.
Some tests, such as those under LibGL, try to load shared libraries at
runtime which are stored in Build/lagom. Because all tests' working
directory is that of their CMakeLists.txt file, they fail to find those
shared libraries at runtime.
Set LD_LIBRARY_PATH to Build/lagom so the tests may find the shared
libraries. This fixes, for example, both of these commands:
./Meta/serenity.sh test lagom TestAPI
./Meta/serenity.sh run lagom TestAPI
When we call add_test() from each test's CMakeLists.txt, we specify the
working directory to be that of the CMakeList.txt file itself. Create a
property to store that directory and reference it when run-lagom-target
is invoked by serenity.sh.
Note that for non-test Lagom targets which do not set this property, the
working directory will be Build/lagom (because the property will be the
empty string, which CMake resolves relative to the current build path).
The current config on GitHub Actions does not use ccache, so it takes
quite a while to build. Instead, let's just run these tests on Azure
where we already build Ladybird and have ccache enabled. This also lets
us sanitize LibWeb on both Linux and macOS.
The script changes here are to A) handle differences between Azure and
GitHub Actions and B) to support running on macOS.
