This is a first port of a simple program to LibMain. A bunch of code is
immediately simplified thanks to the LibSystem wrappers and ability to
use TRY(). This is pretty cool!
This creates an error that contains the name of the syscall that failed.
This allows error handlers to print out the name of the call if they
want to. :^)
By linking with LibMain, your program no longer needs to provide main().
Instead, execution begins in this function:
ErrorOr<int> serenity_main(Main::Arguments);
This allows programs that link with LibMain to use TRY() already in
their entry function, without having to do manual ErrorOr unwrapping.
This is very experimental, but it seems like a nice idea so let's try it
out. :^)
This is a hack to avoid a circular dependency issue with the stack check
failure handler being in LibC.
This is not ideal, and there's most likely a better way to solve this.
That said, LibSystem should not have anything but thin wrappers around
system calls, so stack protectors have limited utility here anyway.
Avoid including a per-translation unit copy of all these functions.
Also, drive-by two clang-tidy fixes for readability-qualified-auto and
readability-implicit-bool-conversion.
This is now as defined in the spec. However since we execute async
functions in bytecode by transforming it to a generator function it must
have a prototype for the GeneratorObject. We check whether it is an
async function and in that case use the hardcoded generator object
prototype. This also ensures that user code cannot override this
property thus preventing exposing internal implementation details.
Removes dotted circle from 1752, 1753, 1E8D0-1E8D6 (apparently these
are combining mark placeholders in the unicode charts and should not
have been drawn).
In particular, malicious programs used to be able to set arbitrary
values as "format", which could cause UB (most likely a crash).
Furthermore, we do not transmit palette data, so an application sending
an indexed bitmap cannot possibly expect the other side to receive a
useful image. Therefore, we refuse to build a bitmap.
The clipboard cannot reasonably contain the empty string. The clipboard
can be empty (i.e. cleared), sure, but that this check was about whether
the clipboard contained the empty string.
This cannot easily happen for two reasons:
- TextEditor GUI elements disable their copy actions when the selection
is empty.
- Clipboard::set_data, through which all text-copying operates,
implicitly forbids empty strings, because Process::sys$anon_create
forbids empty anonymous files.
- Even if it were sent (e.g. by creating a non-empty anonymous file and
sending it manually to the Clipboard server), it would not be
received, because decode(Decoder&, Core::AnonymousBuffer&) goes
through mmap() with a size of 0, which also is forbidden by the
Kernel.
In other words, if the clipboard is never the empty text, therefore
checking this condition is pointless, and we can save a roundtrip to the
Clipboard server.
