The previous version of this was pretty bad and caused a lot of
odd behevaiour to occur. We now abstract a lot of the allocation
behind a `template`d pool class that handles all of the memory
allocation.
The number of UHCI related files is starting to expand to the point
where it's best if we move this into their own subdirectory. It'll
also make it easier to manage when we decide to add some more
controller types (whenever that may be)
This patch prevents CppComprehensionEngine from endlessly looping when
there's a circular #include in the code.
We now keep track of the set of currently processed files, and will not
re-process a file if it already exists in this set.
When we're done with processing a file it is removed from this set.
The pragma once directive is not yet implemented, but regardless a
mechanism that prevents #include loops even in the absence of
pragma once is desired.
This brings `glGetFloatv` more inline with the other `glGet`
functions. We should prevent crashing in the driver as much as
possible and instead let the application deal with the generated
GL error.
Previously debug-kernel.sh wouldn't detect source if you launched the
script from the wrong path. By explicitly setting the name, source will
be loaded by gdb in all situations.
If a member is an empty class, the standard normally stats that it needs
to have a size of at least 1 byte in order to guarantee that the
addresses of distinct objects of the same type are always distinct.
However as of c++20, we can use [[no_unique_address]] to instruct the
compiler that if the member has an empty type, it may optimize it to
occupy no space.
Instead of constructing a String and converting that to a PropertyName
on the fly, we can just leverage CommonPropertyNames, add a couple more
and directly pass ready-to-use PropertyNames with pre-allocated Strings.
Otherwise, we emit a menu_item_left to the WindowServer client even
though the mouse never left the menu item (as is the case when a
disabled menu item is clicked).
Since we operate in screen space where y points down we need to reverse
what is considered clock wise and what is considered counter clockwise.
The rasterizer always expects triangles with a consistent winding order
thus swap 2 vertices if necessary to reverse the winding before passing
the triangle on to the rasterization stage.
The previous clipping implementation was problematic especially when
clipping against the near plane. Triangles are now correctly clipped
using homogenous coordinates against all frustum planes.
Texture coordinates and vertex colors are now correctly interpolated.
The earier implementation was just a placeholder.
GL_VERSION: The spec mandates the following format: x.y or x.y.z
optionally followed by text separated by space.
GL_EXTENSIONS: Return empty string. We do not support any extensions.
glTexCoord should behave like glColor. It only updates a gl context
variable that contains the current texture coordinates. The vertex is
only actually created once glVertex is called.
aplay would previously just retry a couple of times when the loader
encountered an error, even though there is a standard mechanism for
loaders to report errors. Now, aplay outputs these errors to the user
and exits.
FlacLoader initialized, but never used its resampler; this is now fixed
and all subframes are resampled before decorrelation occurs. FLAC files
with non-44100-Hz sample rates now play properly.
Floating-point ratios are inherently imprecise, and can lead to
unpredictable or nondeterministic behavior when resampling and expecting
a certain number of resulting samples. Therefore, the resampler now uses
integer ratios, with almost identical but fully predictable behavior.
This also introduces the reset() function that the FLAC loader will use
in the future.
Since readdir() actually gives us the file type, we don't have
to stat it again to know whether it's a directory! This means that
'find /' can process the majority of files without ever calling
stat() on them, simply by reading directories.
In addition, the TypeCommand (-t) can make use of this info too,
so, for instance, a 'find / -t d' does not need to stat anything
either.
This gives us a final huge speedup :^)
...in a few more places, at least.
find(1) is about to start relying on the reported types more or less
reflecting reality. This is especially relevant for magic symlinks
in ProcFS.
We have multiple commands that are implemented in terms of stat.
Let's cache the stat in FileData after we query it once.
This gives us another large speed-up :^)
While Core::DirIterator is nice C++ API, we want more low-level
control here. In particular, we want to open the directory using
openat(), to also not pay the cost of re-traversing the path
components we have already resolved previously.
This gives us another nice speedup :^)
Also, in the future commits this will allow us to make use of more
data from the returned struct dirent, to speed things up even further.
This speeds things up noticeably :^)
The idea here is that a directory fd is a way to hold onto
the results of path resolution that the kernel has already done
for us. This way we don't ask the kernel to resolve the same
parent directories over and over.
The `-z,text` linker flag causes the linker to reject shared libraries
and PIE executables that have textrels. Our code mostly did not use
these except in one place in LibC, which is changed in this commit.
This makes GNU ld match LLD's behavior, which has this option enabled by
default.
TEXTRELs pose a security risk, as performing these relocations require
executable pages to be written to by the dynamic linker. This can
significantly weaken W^X hardening mitigations.
Note that after this change, TEXTRELs can still be used in ports, as the
dynamic loader code is not changed. There are also uses of it in the
kernel, removing which are outside the scope of this PR. To allow those,
`-z,notext` is added.
