The current implementation fails if a file in the archive is not valid
UTF-8. The CLDR 44.0.1 package unfortunately contains such files (it
errantly has .DS_Store files).
If a unit tests defines a `deps` array, the unit test template would
have tried to overwrite it (and it is actually an error to overwrite
a non-empty list with another non-empty list).
FP contraction is a standard-conforming behavior which allows the
compiler to calculate intermediate results of expressions containing
floating point numbers with a greater precision than the expression type
allows. And in theory, it enables additional optimizations, such as
replacing `a * b + c` with fma(a, b, c).
Unfortunately, it is extremely hard to predict when the contraction will
happen. For example, Clang 17 on x86_64 with the default options will
use FMA only for constant-folded non-constexpr expressions. So, in
practice, FP contraction leads to hard-to-find bugs and inconsistencies
between executables compiled with different toolchains or for different
OSes. And we had two instances of this happening last week.
Since we did not ever used -mfma on x86_64, this patch can only possibly
regress performance on Apple ARM devices, where FMA is enabled by
default. However, this regression will likely be negligible since the
difference would be one additional add instruction, which would be then
likely executed in parallel with something else.
With these changes it is now possible to create OpenGL context on macOS
and run GPU-painter. For now only QT client has a CLI param that turns
it on though.
This is a bit spammy now that we are performing some overload resolution
at build time. The fallback to an interface has generally worked fine on
the types it warns about (BufferSource, Module, etc.) so let's not warn
about it for every build.
Previously, a DNS packet containing an invalid name would be returned
with an empty name. With this change, an error is returned if any error
is encountered during parsing.
Aside from the obvious performance benefits, this will allow us to
properly handle dictionary types. (whose dictionary-ness is only known
at build-time)
Much of the rest of the overload resolution algorithm steps can (and
should) be evaluated at build-time as well, but this is a good first
step.
* Count files that are password-protected
* Use `pdf --json` to groups by missing feature and then by file
instead of by file first, feature second
* Count files that render with issues
* Print number of files without issues last
* Always print all crash stacks
Fix the recursive directory bug in CLion Nova EAP's CMake version in a
way that doesn't also break `./Meta/serenity.sh run lagom ladybird`.
This brings the Lagom minimum required closer to the Serenity minimum
required. Which is still fine, because the serenity.sh script knows how
to build CMake from source if a developer's local copy is too old.
Ran into this with the new EAP for CLion Nova. When using Ladybird as
source directory, we would recursively look in Ladybird --> Lagom -->
Ladybird when exporting components. This seems to be because
ENABLE_LAGOM_LADYBIRD is set to ON by Ladybird's CMakeLists and
something about their build, when invoked from their IDE, has buggy
behavior around the SUBDIRECTORIES directory property.
When wrapping dictionary members, generate_wrap_statement was called
with the pattern "auto {} = ...", where "..." was determined based on
the variable's type. However, in generate_wrap_statement, if a type is
nullable it generates an if statement, so this would end up generating
something along the lines of
if (!retval.member.has_value()) {
auto wrapped_member0_value = JS::js_null();
} else {
auto wrapped_member0_value = JS::Value(...);
}
...which makes the declaration inaccessible. It now generates the same
code, but the "auto" declaration (now an explicit JS::Value declaration)
is outside of the if-statement.
This commit replaces the 5 fuzzers that previously tested LibTextCodec
with a single fuzzer. We now rely on the fuzzer to generate the
encoding and separate it from the encoded data with a magic separator.
This increases the overall coverage of LibTextCodec and eliminates the
possibility of the same error being generated by multiple fuzzers.
When building fuzzers for Oss-Fuzz using `BuildFuzzers.sh --oss-fuzz`,
fuzzer dictionary files are now copied to the `$OUT` directory. This
allows them to be used automatically by the corresponding fuzzer.
LibCore currently cannot depend on LibTimeZone directly. All build-time
code generators depend on LibCore, so there'd be a circular dependency:
LibCore -> LibTimeZone -> GenerateTZData -> LibCore.
So to support parsing time zone names and applying their offsets, add a
couple of weakly-defined helper functions. These work similar to the way
AK::String declares some methods that LibUnicode defines. Any user who
wants to parse time zone names (from outside of LibCore itself) can link
against LibTimeZone to receive full support.
https://cldr.unicode.org/index/downloads/cldr-44
Notable changes that affect us include:
* The Islamic Calendar is now localized as the Hijri Calender (in en-US)
but has not been updated for all locales. So this patch updates tests
where possible and removes a few test cases that currently cannot be
localized.
* The und locale has received more likely subtag data (the und locale is
basically a pseudo-locale meaning "undetermined").
* The exponential symbol in the Arabic number system was changed from
U+0627 to U+0623.
This warning warns about variable-length arrays being a non-standard
extension to the C++ language. We still have a few instances of VLAs, so
let's disable the warning for now.
This does not interfere with `-Wvla`, which we use to completely forbid
this (potentially dangerous) feature in the Kernel and LibCrypto.
The AppKit chrome is now the default, but the Qt chrome may still be
enabled for testing. Let's ensure it can compile in CI, as it has
already broken since the default change.
Previously, some fuzzers were generating an excessive amount of debug
logging. This change explicitly disables debug logging for all fuzzers.
This allows higher test throughput and makes the logs easier to read
when fuzzing locally.
The previous implementation was calling `backtrace()` for every
function call, which is quite slow.
Instead, this implementation provides VM::stack_trace() which unwinds
the native stack, maps it through NativeExecutable::get_source_range
and combines it with source ranges from interpreted call frames.
