Currently, invoking StringBuilder::to_string will re-allocate the string
data to construct the String. This is wasteful both in terms of memory
and speed.
The goal here is to simply hand the string buffer over to String, and
let String take ownership of that buffer. To do this, StringBuilder must
have the same memory layout as Detail::StringData. This layout is just
the members of the StringData class followed by the string itself.
So when a StringBuilder is created, we reserve sizeof(StringData) bytes
at the front of the buffer. StringData can then construct itself into
the buffer with placement new.
Things to note:
* StringData must now be aware of the actual capacity of its buffer, as
that can be larger than the string size.
* We must take care not to pass ownership of inlined string buffers, as
these live on the stack.
This is done by providing Traits<ByteBuffer>::equals functions for
(Readonly)Bytes, as the base GenericTraits<T>::equals is unable to
convert the ByteBuffer to (Readonly)Bytes to then use Span::operator==
This allows us to check if a Vector<ByteBuffer> contains a
(Readonly)Bytes without having to making a copy of it into a ByteBuffer
first. The initial use of this is in LibWeb with CORS-preflight, where
we check the split contents of the Access-Control headers with
Fetch::Infrastructure::Request::method() and static StringViews
such as "*"sv.bytes().
C++20 can automatically synthesize `operator!=` from `operator==`, so
there is no point in writing such functions by hand if all they do is
call through to `operator==`.
This fixes a compile error with compilers that implement P2468 (Clang
16 currently). This paper restores the C++17 behavior that if both
`T::operator==(U)` and `T::operator!=(U)` exist, `U == T` won't be
rewritten in reverse to call `T::operator==(U)`. Removing `!=` operators
makes the rewriting possible again.
See https://reviews.llvm.org/D134529#3853062
ByteBuffer::get_bytes_for_writing() was only ensuring capacity before
this patch. The method needs to call resize to register the appended
data, otherwise it will be overwritten with next data addition.
The compiler would complain about `__builtin_memcpy` in ByteBuffer::copy
writing out of bounds, as it isn't able to deduce the invariant that the
inline buffer is only used when the requested size is smaller than the
inline capacity.
The other change is more bizarre. If the destructor's declaration
exists, gcc complains about a `delete` operation causing an
out-of-bounds array access.
error: array subscript 'DHCPv4Client::__as_base [0]' is partly outside
array bounds of 'unsigned char [8]' [-Werror=array-bounds]
14 | ~DHCPv4Client() = default;
| ^
This looks like a compiler bug, and I'll report it if I find a suitable
reduced reproducer.
Apologies for the enormous commit, but I don't see a way to split this
up nicely. In the vast majority of cases it's a simple change. A few
extra places can use TRY instead of manual error checking though. :^)
This is useful for writing new data at the end of a ByteBuffer. For
instance, with the Stream API:
auto pending_bytes = TRY(stream.pending_bytes());
auto receive_buffer = TRY(buffer.get_bytes_for_writing(
pending_bytes));
TRY(stream.read(receive_buffer));
Same as Vector, ByteBuffer now also signals allocation failure by
returning an ENOMEM Error instead of a bool, allowing us to use the
TRY() and MUST() patterns.
This class is the only reason we have to support krealloc() in the
kernel heap, something which adds a lot of complexity.
Let's move towards a simpler path and do malloc+memset in the
ByteBuffer code (where we know the sizes anyway.)
Other software might not expect these to be defined and behave
differently if they _are_ defined, e.g. scummvm which checks if
the TODO macro is defined and fails to build if it is.
This allows us to mark the slow part (i.e. where we copy the buffer) as
NEVER_INLINE because this should almost never get called and therefore
should also not get inlined into callers.
Previously ByteBuffer::grow() behaved like Vector<T>::resize().
However the function name was somewhat ambiguous - and so this patch
updates ByteBuffer to behave more like Vector<T> by replacing grow()
with resize() and adding an ensure_capacity() method.
This also lets the user change the buffer's capacity without affecting
the size which was not previously possible.
Additionally this patch makes the capacity() method public (again).
Previously GCC came to the conclusion that we were reading
m_outline_capacity via ByteBuffer(ByteBuffer const&) -> grow()
-> capacity() even though that could never be the case because
m_size is 0 at that point which means we have an inline buffer
and capacity() would return inline_capacity in that case without
reading m_outline_capacity.
This makes GCC inline parts of the grow() function into the
ByteBuffer copy constructor which seems sufficient for GCC to
realize that m_outline_capacity isn't actually being read.
When compiling the Kernel with Og, the compiler complains that
m_outline_capacity might be uninitialized when calling capacity()
Note that this fix is not really what we want. Ideally only outline
buffer and outline capacity would need initialized, not the entire
inline buffer. However, clang considers the class to not be
default-constructible if we make that change, while gcc accepts it.
Previously ByteBuffer would internally hold a RefPtr to the byte
buffer and would behave like a reference type, i.e. copying a
ByteBuffer would not create a duplicate byte buffer, but rather
two objects which refer to the same internal buffer.
This also changes ByteBuffer so that it has some internal capacity
much like the Vector<T> type. Unlike Vector<T> however a byte
buffer's data may be uninitialized.
With this commit ByteBuffer makes use of the kmalloc_good_size()
API to pick an optimal allocation size for its internal buffer.
Creating a ByteBuffer involves two allocations:
-One for the ByteBufferImpl object
-Another one for the actual byte buffer
This changes the ByteBuffer and ByteBufferImpl classes
so only one allocation is necessary.
