When trying to figure out the correct implementation, we now have a very
strong distinction on plugins that are well suited for sniffing, and
plugins that need a MIME type to be chosen.
Instead of having multiple calls to non-static virtual sniff methods for
each Image decoding plugin, we have 2 static methods for each
implementation:
1. The sniff method, which in contrast to the old method, gets a
ReadonlyBytes parameter and ensures we can figure out the result
with zero heap allocations for most implementations.
2. The create method, which just creates a new instance so we don't
expose the constructor to everyone anymore.
In addition to that, we have a new virtual method called initialize,
which has a per-implementation initialization pattern to actually ensure
each implementation can construct a decoder object, and then have a
correct context being applied to it for the actual decoding.
These instances were detected by searching for files that include
Array.h, but don't match the regex:
\\b(Array(?!\.h>)|iota_array|integer_sequence_generate_array)\\b
These are the three symbols defined by Array.h.
In theory, one might use LibCPP to detect things like this
automatically, but let's do this one step after another.
The order of PNG compression is raw pixel data -> filter -> compress.
For decompression, the order is reversed, so that means uncompress ->
unfilter -> raw pixel data. Previously, the PNG decoder was converting
to raw pixel data before unfiltering, which was a problem when using
indexed color palettes, since each pixel's palette index could change
during unfiltering (e.g. it was unfiltering after already choosing
the color from the palette index). This was leading to 'Palette index
out of range' errors on files that both:
- Had scanlines with some sort of filtering
- Didn't use the full range of possible palette indices for their bit
depth.
Also, because filtering now happens before converting to pixel data,
filtering acts on bytes instead of pixels, meaning that the
implementation of each filter type now maps much more directly to
the specification:
http://www.libpng.org/pub/png/spec/1.2/PNG-Filters.html
Error::from_string_literal now takes direct char const*s, while
Error::from_string_view does what Error::from_string_literal used to do:
taking StringViews. This change will remove the need to insert `sv`
after error strings when returning string literal errors once
StringView(char const*) is removed.
No functional changes.
`static const` variables can be computed and initialized at run-time
during initialization or the first time a function is called. Change
them to `static constexpr` to ensure they are computed at
compile-time.
This allows some removal of `strlen` because the length of the
`StringView` can be used which is pre-computed at compile-time.
The ARGB32 typedef is used for 32-bit #AARRGGBB quadruplets. As such,
the name RGBA32 was misleading, so let's call it ARGB32 instead.
Since endianness is a thing, let's not encode any assumptions about byte
order in the name of this type. ARGB32 is basically a "machine word"
of color.
For PNG color type 2 (TrueColor), if there's a tRNS chunk, it specifies
an R/G/B triplet in the image bit depth. This triplet acts as the
transparency value, and should yield transparent pixels wherever that
specific color value is present.
We now support this. :^)
The tRNS metadata is valid for color types 0, 2 and 3, so let's collect
it in each of those cases.
This ensures that we produce a bitmap with an alpha channel if needed.
We had a bunch of old unused wrapper functions for each image codec that
would load a supported image with a given path. Nobody actually used
them, so let's just get rid of load_png(), load_gif(), etc.
To transparently support multi-frame images, all decoder plugins have
already been updated to return their only bitmap for frame(0).
This patch completes the remaining cleanup work by removing the
ImageDecoder::bitmap() API and having all clients call frame() instead.
Making a bitmap non-volatile after being volatile may fail to allocate
physical pages after the kernel stole the old pages in a purge.
This is different from the pages being purged, but reallocated. In that
case, they are simply replaced with zero-fill-on-demand pages as if
they were freshly allocated.
This was a really weird thing to begin with, purgeable bitmaps were
basically regular bitmaps without a physical memory reservation.
Since all the clients of this code ended up populating the bitmaps
with pixels immediately after allocating them anyway, there was no
need to avoid the reservation.
Instead, all Gfx::Bitmaps are now purgeable, in the sense that they
can be marked as volatile or non-volatile.
The only difference here is that allocation failure is surfaced when
we try to create the bitmap instead of during the handling of a
subsequent page fault.
AK's version should see better inlining behaviors, than the LibM one.
We avoid mixed usage for now though.
Also clean up some stale math includes and improper floatingpoint usage.
Problem:
- `static` variables consume memory and sometimes are less
optimizable.
- `static const` variables can be `constexpr`, usually.
- `static` function-local variables require an initialization check
every time the function is run.
Solution:
- If a global `static` variable is only used in a single function then
move it into the function and make it non-`static` and `constexpr`.
- Make all global `static` variables `constexpr` instead of `const`.
- Change function-local `static const[expr]` variables to be just
`constexpr`.
This wasn't much of a problem before because copying the ByteBuffer
merely copied the RefPtr but now that ByteBuffer behaves like Vector
this causes unnecessary allocations.
SPDX License Identifiers are a more compact / standardized
way of representing file license information.
See: https://spdx.dev/resources/use/#identifiers
This was done with the `ambr` search and replace tool.
ambr --no-parent-ignore --key-from-file --rep-from-file key.txt rep.txt *
As per the PNG specification: "Filter method is a single-byte
integer that indicates the preprocessing method applied to the
image data before compression. At present, only filter method 0
(adaptive filtering with five basic filter types) is defined."
As per the PNG specification: "For all x < 0, assume Raw(x) = 0 and
Prior(x) = 0. On the first scanline of an image (or of a pass of an
interlaced image), assume Prior(x) = 0 for all x."
PNGs use deflate with zlib, however we were just skipping the zlib
bytes and then piping it into deflate decompressor. Since we have a
zlib decompressor, lets use that instead.
This has the added benefit of extra error checking.
The previous names (RGBA32 and RGB32) were misleading since that's not
the actual byte order in memory. The new names reflect exactly how the
color values get laid out in bitmap data.
This commit removes the only 3rd party library (and its usages)
in serenity: puff, which is used for deflate decompression. and
replaces it with the existing original serenity implementation
in LibCompress. :^)
(...and ASSERT_NOT_REACHED => VERIFY_NOT_REACHED)
Since all of these checks are done in release builds as well,
let's rename them to VERIFY to prevent confusion, as everyone is
used to assertions being compiled out in release.
We can introduce a new ASSERT macro that is specifically for debug
checks, but I'm doing this wholesale conversion first since we've
accumulated thousands of these already, and it's not immediately
obvious which ones are suitable for ASSERT.
