Fixes bug when "clip" property does not affect abspos children.
This change makes "clip" property to be applied together with
"overflow: hidden" in `apply_clip_overflow_rect()` that already
handles abspos children correctly.
This matches if the element has a placeholder, and that placeholder is
currently visible. This applies to `<input>` and `<textarea>` elements,
but our `<textarea>` is very limited so does not support placeholders.
These APIs only perform small allocations, and are only used by LibJS
and the time zone settings widget. Callers which could only have failed
from these APIs are also made to be infallible here.
These APIs only perform small allocations, and are only used by LibJS.
Callers which could only have failed from these APIs are also made to
be infallible here.
Grid specification https://www.w3.org/TR/css-grid-2/#z-order defines
special painting order for grid items which should be the same as for
defined for inline-blocks in CSS2.
Check the width of the next token after white space to decide line
breaks. The next width can also be the total width of multiple tokens.
This better follows the CSS Text specification and matches behavior of
other browsers.
Fixes#20388.
We still don't know how to resolve font-relative lengths in <img sizes>
since we don't always have font size information available at this stage
in the pipeline, but we can at least handle viewport-relative lengths.
This fixes an issue on many websites where low-resolution images were
loaded (appropriate for a small viewport) even when the viewport is big.
And treat them as "auto" for now, per CSS-SIZING-3, with a FIXME about
supporting more layout directions.
This fixes an issue on MDN where `height: max-content` was not
overriding height from non-CSS presentational hints.
CSS-ALIGN-3 tells us that `normal` behavior inside flex containers is
simply to behave as `stretch` so this patch makes them behave the same
inside FFC.
Furthermore, we change the `align-items` initial value to `normal`,
matching other engines.
I'm about to make StackingContext traverse the paintable tree instead of
actually traversing the layout tree, and it turns out we were not
creating paintables for these SVG elements.
Also switch them to Layout::Box instead of the default InlineNode to
make the trees look a bit less weird. Ultimately, we should do something
specialized for these subtrees, but for now this'll do.
This patch just adds the new root paintable and updates the tests
expectations. The next patch will move painting logic from the layout
viewport to the paint viewport.
Instead of assuming that the WebDriver binary is in PATH or the two
most common build directories for our scripts, add a command line
argument to the script to pass a WebDriver binary.
We should be documenting required pacakges elsewhere and installing them
in the setup step of CI.
This also fixes a problem where the run script would fail if you already
had a cloned wpt directory.
The setup is a bit peculiar: both the definition and the use site of
these TLS variables have to be in a shared library, otherwise the linker
might relax the global-dynamic access mode to something that doesn't
require a `__tls_get_addr` call.
This has to cheat and use a screenshot but thanks to the "Take Full
Screenshot" feature of Ladybird, it is very easy to update this test.
The steps are documented in the test.
When a box is sized under max-content constraint, any percentage value
set for max-width should be considered as if it were infinite. In other
words, it should have no effect on restricting the box's width.
This doesn't seem to actually have fixed any bugs, as having
FillOpacity instead of StrokeOpacity in the call to parse_css_value
doesn't seem to have actually been causing bugs. But, I still think it's
worthwhile correcting.
The reason that it wasn't causing bugs is that having FillOpacity
instead of StrokeOpacity in the call to parse_css_value means that when
parsing the value is compared to the acceptable values for that property
(for example the value can only be a percentage, or a number, etc.). In
this case both FillOpacity and StrokeOpacity seem to accept the same
values.
The SVG G container should have the same size as its children. This
fixes a bug when there was an opacity value on the G element, as in
StackingContext it would try and get a bitmap of the element which would
be empty due to it having no size.
We now keep the color value as a StyleValue up until we go to paint the
gradient, which makes `currentColor` work, along with any other color
values that can't be immediately converted into a `Gfx::Color` while
parsing.
If we run an inline script from the HTML parser, it may append a text
node to the current insertion point.
If there was text content immediately following the script element,
we would previously overwrite the script-inserted text content, due to
an oversight in the way we select an appropriate insertion point
This patch fixes the issue by only inserting parser content into
existing text nodes if they are empty.
Instead of applying relative offsets (like position:relative insets)
during painting and hit testing, we now do a pass at the end of layout
and assign the final resolved offsets to paintables.
This makes painting and hit testing easier since they don't have to
think about relative offsets, and it also fixes a bug where offsets were
not applied to text fragments inside inline-flow elements that were
themselves position:relative.
The ref tests runner takes screenshots of both the input page and the
expected page, then compares them. Ref testing allows us to catch
painting bugs, which cannot be detected with the layout and text tests
we already have.
With ref tests, we'll likely want to reuse the same expectation page
for multiple inputs. Therefore, there's a `manifest.json` file that
describes the relationship between inputs and expected outputs.
Stop worrying about tiny OOMs. Work towards #20449.
While going through these, I also changed the function signature in many
places where returning ThrowCompletionOr<T> is no longer necessary.
Just like with input buffered streams, we don't currently have a use
case for output buffered streams which aren't seekable, since the main
application are files.
Changing `calculate_min_content_heigh()` and
`calculate_min_content_heigh()` to accept width as `CSSPixels`, instead
of `AvailableSize` that might be indefinite, makes it more explicit
that width is supposed to be known by the time height is measured.
This change has a bit of collateral damage which is rows height
calculation regression in `table/inline-table-width` that worked before
by accident.
We now apply MathML's default user agent style sheet along with other
default styles. This sheet is not mixed in with the other styles in
CSS/Default.css because it is a namespaced stylesheet and so has to
be its own sheet.
Using avilable space directly while resolving table container width
allows to avoid assigning it to table wrapper box content width which
sometimes involves infinite (saturated) values.
Also this allows to get rid of set_max_content_width() which is a hack
that allows to bypass set_content_width() to assign infinite
(saturated) width to a box.
Closes https://github.com/SerenityOS/serenity/issues/19521
Replicate the more conservative way it's done for other nodes, for
which we verify whether they have a paintable before doing
painting-related operations with it.
Fixes crash on https://www.haiku-os.org/.
Returning greatest_child_width() from automatic_content_width() in BFC
if root box children are inline and there are min/max-width that caused
width to be changed after IFC layout while content_width should be
always set to correct value by layout_inline_children() regardless of
layout mode.
Using the kernel stack is preferable, especially when the examined
strings should be limited to a reasonable length.
This is a small improvement, because if we don't actually move these
strings then we don't need to own heap allocations for them during the
syscall handler function scope.
In addition to that, some kernel strings are known to be limited, like
the hostname string, for these strings we also can use FixedStringBuffer
to store and copy to and from these buffers, without using any heap
allocations at all.
I'm leaving the --use-bytecode CLI option here as a no-op for now, until
we get all the scripts updated. But the program always runs in bytecode
mode now.
- Out-of-flow items should not affect grid layout
- "The static position of an absolutely-positioned child of a grid
container is determined as if it were the sole grid item in a grid
area whose edges coincide with the content edges of the grid
container."
Now that ""_string is infallible, the only benefit of explicitly
constructing a short string is the ability to do it at compile-time. But
we never do that, so let's simplify the API and remove this
implementation detail from it.
The features provided by these classes should be used eventually, but so
far we've been maintaining these classes for over 2 years without any
actual use. We can restore them when it comes time to actually use them.
This will ensure that we don't leak any memory while playing back
audio.
There is an expectation value in the test that is only set to true when
PulseAudio is present for the moment. When any new implementation is
added for other libraries/platforms, we should hopefully get a CI
failure due to unexpected success in creating the `PlaybackStream`.
To ensure that we clean up our PulseAudio connection whenever audio
output is not needed, add `PulseAudioContext::weak_instance()` to allow
us to check whether an instance exists without creating one.
This object is available as `window.internals` (or just `internals`) and
is only accessible while running in "test mode".
This first version only has one API: gc(), which triggers a garbage
collection immediately.
In the future, we can add more APIs here to help us test parts of the
engine that are hard or impossible to reach via public web APIs.
This could happen if a sequence of '0' parts was followed by a longer
sequence of '0' parts at the end of the host. The first sequence was
being used for the compress, and not the second.
For example, [1:1:0:0:1:0:0:0] was being serialized as: [1:1::1:0:0:0]
instead of [1:1:0:0:1::].
Fix this by checking at the end of the loop if we are in the middle of a
sequence of '0' parts that is longer than the current longest.
We were already parsing non-function-syntax :host, so let's also do
the :host(...) variant. Note that we don't have matching for these yet.
This fixes many issues on sites generated by Wix, as they often have
selector lists that include some :host() selector, and we'd reject the
entire rule after failing to parse it.
This stuff is pretty hairy since the specifications don't give any
guidance on which widths to use when calculating the intrinsic height of
flex items in a column layout.
However, our old behavior of "treat anything indefinite as fit-content"
was definitely not good enough, so this patch improves the situation by
considering values like `min-content`, `max-content` and `fit-content`
separately from `auto`, and making the whole flex layout pipeline aware
of them (in the cross axis context).
Use the max-width of percentage cells instead of min-width as the
reference to be used to compute the total table width. The specification
only suggests that the UA should try to satisfy percentage constraints
and this behavior is more consistent with other browsers.
Since the existing Promise class is designed with deferred tasks on the
main thread only, we need a new class that will ensure we can handle
promises that are resolved/rejected off the main thread.
This new class ensures that the callbacks are only called on the same
thread that the promise is fulfilled from. If the callbacks are not set
before the thread tries to fulfill the promise, it will spin until they
are so that they will run on that thread.
Fixes infinite spinning in the cases when CSSPixels does not have
enough precision to represent increase per track which happens when
very small extra_space got divided by affected tracks number.
Change associativity in computing of replaced element size to improve
precision of division.
Fixes vertically squashed image from Mozilla splash page MDN example.
This patch implements "Overflow Viewport Propagation" from CSS-OVERFLOW.
It fixes an issue where many websites were not scrollable because they
had `overflow: scroll` on the body element and we didn't propagate it.
Pseudo-elements like ::before and ::after were discarded when their
content property was an empty string (ignoring whitespace), because they
are anonymous containers with no lines.
Our previous way around it was to add an empty line box (see b062a0fb7c)
however it didn't actually work for cases described in the previous
commit.
This makes avatars and cover arts square on last.fm and "fixes" the test
css-pseudo-element-should-not-be-affected-by-presentational-hints.html.
Unfortunately, this also regresses on block-and-inline/clearfix.html,
but that hopefully will be handled in subsequent commit.
Properties like min-width, max-width, etc, should be ignored while we're
trying to determine the intrinsic size of a flex container.
This fixes an infinite recursion when using an intrinsic size keyword as
the max-width of a flex column container.
Note that this behavior is marked as AD-HOC in code comments because
specs don't tell us how to achieve intrinsic sizing.
We can now load product pages on the Twinings site, such as
https://twinings.co.uk/products/earl-grey-100-tea-bags :^)
Fixes the issue that before "automatic minimum size" were used to size
flexible tracks even though specification says is should be "minimum
contribution"
Parsing 'data:' URLs took it's own route. It never set standard URL
fields like path, query or fragment (except for scheme) and instead
gave us separate methods called `data_payload()`, `data_mime_type()`,
and `data_payload_is_base64()`.
Because parsing 'data:' didn't use standard fields, running the
following JS code:
new URL('#a', 'data:text/plain,hello').toString()
not only cleared the path as URLParser doesn't check for data from
data_payload() function (making the result be 'data:#a'), but it also
crashes the program because we forbid having an empty MIME type when we
serialize to string.
With this change, 'data:' URLs will be parsed like every other URLs.
To decode the 'data:' URL contents, one needs to call process_data_url()
on a URL, which will return a struct containing MIME type with already
decoded data! :^)
By not clearing the buffer, we were leaking the path part of a URL into
the query for URLs without an authority component (no '//host').
This could be seen most noticeably in mailto: URLs with header fields
set, as the query part of `mailto:user@example.com?subject=test` was
parsed to `user@example.comsubject=test`.
data: URLs didn't have this problem, because we have a special case for
parsing them.
When toggling `display: none` on an element, it can go from having a
layout subtree to not having one. In the `none` case, we were previously
leaving stale layout nodes hanging off DOM nodes in the subtree.
These layout nodes could be queried for outdated information and
probably other things that we shouldn't allow.
Fix this by having TreeBuilder prune any old layout nodes hanging off
nodes in a subtree after its subtree root doesn't produce a layout node.
The first implementation of this property was just plain wrong. Looks
like this property isn't used a lot as I found the issue by reviewing
the code and not because of a specific image.
The test image is a 32x32 mosaic of alternating black and yellow pixels,
it was generated using this code:
Bitdepth 8
RCT 1
Width 32
Height 32
if W-WW-NW+NWW > -300
- Set -1000
- Set 900
This is defined in the spec, but was missing in our table. Fix this, and
add a spec comment for what is missing. Also begin a basic text based
test for URL, so we can get some coverage of LibWeb's usage of URL too.
This takes the previous alternation optimisation and applies it to all
the alternation blocks instead of just the few instructions at the
start.
By generating a trie of instructions, all logically equivalent
instructions will be consolidated into a single node, allowing the
engine to avoid checking the same thing multiple times.
For instance, given the pattern /abc|ac|ab/, this optimisation would
generate the following tree:
- a
| - b
| | - c
| | | - <accept>
| | - <accept>
| - c
| | - <accept>
which will attempt to match 'a' or 'b' only once, and would also limit
the number of backtrackings performed in case alternatives fails to
match.
This optimisation is currently gated behind a simple cost model that
estimates the number of instructions generated, which is pessimistic for
small patterns, though the change in performance in such patterns is not
particularly large.
In order to follow spec text to achieve this, we need to change the
underlying representation of a host in AK::URL to deserialized format.
Before this, we were parsing the host and then immediately serializing
it again.
Making that change resulted in a whole bunch of fallout.
After this change, callers can access the serialized data through
this concept-host-serializer. The functional end result of this
change is that IPv6 hosts are now correctly serialized to be
surrounded with '[' and ']'.
Previously, we always rounded border-widths up when converting them to
device pixels. However, the spec asks us to follow a specific algorithm
to "snap" these values, so that the computed value is snapped.
The difference from before, is that widths of between 0 and 1 device
pixels are rounded up to 1, and and values larger than 1 are rounded
down.
If a math function resolves to `<length>` or `<percentage>`, then it
will by definition also resolve to `<length-percentage>`. (Same for any
other basic types.) Since we were checking `<length-percentage>` first
and then bailing if no given properties could accept that, math
functions would always fail to match a property that just accepts a non
`-percentage` type.
Multiple patches may be concatenated in the same patch file, such as git
commits which are changing multiple files at the same time. To handle
this, parse each patch in order in the patch file, and apply each patch
sequentially.
To determine whether we are at the end of a patch (and not just parsing
another hunk) the parser will look for a leading '@@ ' after every hunk.
If that is found, there is another hunk. Otherwise, we must be at the
end of this patch.
Previously patch would always expect the file that it was patching to
exist (even it were empty). If we know that the patch is creating a file
from nothing (i.e has a start line of '0'), then we treat a file that
doesn't exist as if it has no content lines.
Implement the patch '-p' / '--strip' option, which strips the given
number of leading components from filenames parsed in the patch header.
If not given this option defaults to the basename of that path.
This is a universal value like `initial` and `inherit` and works by
reverting the current value to whatever we had at the start of the
current cascade origin.
The implementation is somewhat inefficient as we make a copy of all
current values at the start of each origin. I'm sure we can come up with
a way to make this faster eventually.
Allow the left margin of a box which creates a block formatting context
to overlap with left floating boxes which are siblings in the document
tree.
Fixes#20233 and the comment layout on https://lobste.rs.
This change makes tree builder omit elements with "display: contents"
from the layout tree during construction. Their child elements are
instead directly appended to the parent element in layout tree.
Similar to commit 0652cc4, we now generate 2-stage lookup tables for
case conversion information. Only about 1500 code points are actually
cased. This means that case information is rather highly compressible,
as the blocks we break the code points into will generally all have no
casing information at all.
In total, this change:
* Does not change the size of libunicode.so (which is nice because,
generally, the 2-stage lookup tables are expected to trade a bit
of size for performance).
* Reduces the runtime of the new benchmark test case added here from
1.383s to 1.127s (about an 18.5% improvement).
Currently, the `isobmff` utility will only print the media file type
info from the FileTypeBox (major brand and compatible brands), as well
as the names and sizes of top-level boxes.
For malformed tables which only have cells with span greater than 1, the
content sizes for row and column aren't initialized to non-zero values.
Avoid undefined behavior in such cases, which sometimes show up on
Wikipedia.
This image is exactly the same as the previous one, excepted the RCT
transformation. It has been generated with:
Width 64
Height 64
RCT 29
Upsample 2
Bitdepth 10
if N > 300
- NE -6
- W 6
We currently produce a single table for all categories of code point
properties (GeneralCategory, Script, etc.). Each row contains a field
indicating the range of code points to which that property applies. At
runtime, we then do a binary search through that table to decide if a
code point has a property.
This changes our approach to generate a 2-stage lookup table for each of
those categories. There is an in-depth explanation of these tables above
the new `create_code_point_tables` method. The end effect is that code
point property lookup is reduced from a binary search to constant-time
array lookups.
In total, this change:
* Increases the size of libunicode.so from 2.7 MB to 2.9 MB.
* Reduces the runtime of the new benchmark test case added here from
3.576s to 1.020s (a 3.5x speedup).
* In a profile of resizing a TextEditor window with a 3MB file open,
the runtime of checking if a code point has a word break property
reduces from ~81% to ~56%.
We started generating this data in commit 0505e03, but it was unused.
It's still not used, so let's remove it, rather than bloating the size
of libunicode.so with unused data. If we need it in the future, it's
trivial to add back.
Note we *have* always used the block name data from that commit, and
that is still present here.
Auto margins used together with justify-content would previously
result in children being positioned outside their parent. This was
solved by letting auto margins take precedence when they are used,
which was already implemented to some extent before, but not
fully.
Containers with both flex reverse and justify content would
sometimes place children outside the container. This happened
because it assumed any reversed container would have items
aligned to the right, which isn't true when using eg. `flex-end`.
Both `justify-content: start` and `justify-content: end` are now
also independent of the reverseness.
After switching to fixed-point arithmetic in CSSPixels, it no longer
supports representing infinite values, which was previously the case
for remaining_free_space in FFC. Using Optional that is not empty only
when value is finite to store remaining_free_space ensures that
infinity is avoided in layout calculations.
This fixes an assertion on https://amazon.com/ since WindowProxy
would advertise "0" as an own property key, but then act like it was
a bogus property when actually queried for it directly.
Before this change, we always derived a box's baseline from its last
child, even if the last child didn't have any line boxes inside.
This caused baselines to slip further down vertically than expected.
There are more baseline alignment issues to fix, but this one was
responsible for a fair chunk of trouble. :^)
I misunderstood the spec step for checking whether the host 'ends with a
number'. We can't simply check for it if ends with a number, this check
is actually an algorithm which is required to avoid detecting hosts that
end with a number from an IPv4 host.
Implement this missing step, and add a test to cover this.
After the CSSPixels implementation evolved from a wrapper of double
to a fixed-point saturated math arithmetic implementation, it makes
sense to have separate tests for it.
Using fixed-point saturated arithmetics for CSSPixels allows to avoid
accumulating floating-point errors.
This implementation is not complete yet: currently saturated
arithmetics implemented only for addition. But it is enough to not
regress any of layout tests we have :)
See https://github.com/SerenityOS/serenity/issues/18566
Returning a reference resulted in Mail's use of Promise causing a crash.
Also, awaiting an already-awaited promise is an odd thing to do anyway,
so let's just make it release the resolved/rejected value instead of
returning a reference to it.
Co-Authored-By: Valtteri Koskivuori <vkoskiv@gmail.com>
Follow the computing column measures section of the specification, which
gives an algorithm for setting intrinsic percentage widths when spanning
columns are involved.
This is intended to be used in adding test-js tests where there is
different behaviour between the AST interpreter and bytecode mode.
In particular, this is useful for tests which fail in AST, but pass in
bytecode, as the AST interpreter is run in CI but bytecode is not.
This image uses the modular encoding with a very simple prediction tree.
It also makes use of two features: upsampling (x2 factor) and a
non-standard bit depth (10 bits). The file has been generated on
https://jxl-art.surma.technology/ , with the following input:
Width 64
Height 64
Upsample 2
Bitdepth 10
if N > 300
- NE -6
- W 6
Change how we store type of columns. It was used where the specification
only distinguishes between percent and everything else, so it makes more
sense to store and use it as a boolean.
The specification says we should distribute excess width proportionally
to the width of the cell, not to the preferred increment. Doing the
latter leads to distributing all excess width to just the cells which
demand some increment, even if it's very modest. Moreover, there's code
which partially implements the correct criteria just below the one we
remove here.
These passes have not been shown to actually optimize any JS, and tests
have become very flaky with optimizations enabled. Until some measurable
benefit is shown, remove the optimization passes to reduce overhead of
maintaining bytecode operations and to reduce CI churn. The framework
for optimizations will live on in git history, and can be restored once
proven useful.
The Heap::uproot_cell() API was used to implement markAsGarbage() which
was used in 3 tests to forcibly destroy a value, even if it had
references on the stack or elsewhere.
This patch rewrites the 3 tests that used this mechanism to be
structured in a way that allows garbage collection to collect the values
as intended without hacks. And now that the uprooting mechanism is no
longer needed, it's uprooted as well.
This fixes 3 test-js tests in bytecode mode. :^)
I added this file thinking it was necessary for the wpt run command.
However, it's only needed for updating expectations metadata. Since wpt
run always regenerates MANIFEST.json before updating expectations, we
can safely delete this file from the repository.
Reordering these calls allow us to ensure that all encoders are able to
return the size of the image before they are requested to decode the
whole bitmap.
The Test262RunnerHandler class in test-test262 was made in order to
spawn a subprocess, connect to its input/output error pipes, and obtain
its return value.
Later on, a copy of this implementation was added to TestSed with
modifications, such as adding support for reading from the output pipes
as well.
Unify these two implementations into a new Core::Command class. This new
implementation is more closely modeled from the TestSed implementation
due to the extra functionality it implemented.
Once we've resolved the used flex item width & height, we should allow
percentage flex item sizes to resolve against them instead of forcing
flex items to always treat percentages as auto while doing intrinsic
sizing layout.
Regressed in 8dd489da61.
When specifying either `background-position-x: right` or
`background-position-y: bottom` without an offset value no
EdgeStyleValue was created.
However, the spec says the offset should be optional.
Now, if you do not provide an offset, it creates the EdgeStyleValue
with a default offset of 0 pixels.
We do this by piggybacking on FormattingContext helpers instead of
reinventing the wheel in FlexFormattingContext.
This fixes an issue where `min-width: fit-content` (and other
layout-dependent values) were treated as 0 on flex items.
This makes the cookie banners look okay on https://microsoft.com/ :^)
This is just a straight (and fairly inefficient) implementation of IPv6
parsing and serialization from the URL spec.
Note that we don't use AK::IPv6Address here because the URL spec
requires a specific serialization behavior.
If an inline-block has a percentage height that relies on the auto
height of the containing block, it should always resolve to the
automatic height of the box, regardless of the percentage value. This
change may seem confusing, but it aligns with the behavior of other
engines.
Unlike all other primitives elliptical arcs are non-trivial to
manipulate, it's tricky to correctly apply a Gfx::AffineTransform to
them. Prior to this change, Path::copy_transformed() was still
incorrectly applying transforms such as flips and skews to arcs.
This patch very closely approximates arcs with cubic beziers (I can not
visually spot any differences), which can then be easily and correctly
transformed in all cases.
Most of the maths here was taken from:
https://mortoray.com/rendering-an-svg-elliptical-arc-as-bezier-curves/
(which came from https://www.joecridge.me/content/pdf/bezier-arcs.pdf,
now a dead link).
This tests that we can successfully parse the "everything" TVG files,
which make use of every feature in TinyVG.
Test files taken from https://github.com/TinyVG/examples (MIT).
This test proves the ability of TransformStream to execute
caller supplied code in the flush callback, and have access to
TransformStreamDefaultController.
This test proves the ability of TransformStream to execute
caller supplied code in the start callback, and have access to
TransformStreamDefaultController.
This test proves the ability of TransformStream to execute to execute
caller supplied code in the transform callback that can transform
incoming chunks, and have access to TransformStreamDefaultController.
There are two parts to this fix:
- First, StyleProperties::transformations() would previously omit calc()
values entirely when returning the list of transformations. This was
very confusing to StackingContext which then tried to index into the
list based on faulty assumptions. Fix this by emitting calc values.
- Second, StackingContext::get_transformation_matrix() now always calls
resolve() on length-percentages. This takes care of actually resolving
calc() values. If no reference value for percentages is provided, we
default to 0px.
This stops LibWeb from asserting on websites with calc() in transform
values, such as https://qt.io/ :^)
This change implements following paragraph from placement algorithm in
the spec:
"If the largest column span among all the items without a definite
column position is larger than the width of the implicit grid, add
columns to the end of the implicit grid to accommodate that column
span."
There were places in the grid implementation code with copies of this
text, but those were completely unrelated to the code where they were
being pasted so I removed them.
There were two problems:
1. parse_compressed_object_with_index() parses indirect objects
without going through Parser::parse_indirect_value(), so
push_reference() / pop_reference() weren't called.
Manually call them, both for the indirect object containing
the object stream and for the indirect object within the
object stream.
2. The indirect object within the object stream got decrypted
twice: Once when the object stream data itself got decrypted,
and then incorrectly a second time when the object data within
the stream was read. To fix, disable encryption while parsing
object stream data (since it's already decrypted).
The test is from http://opf-labs.org/format-corpus/pdfCabinetOfHorrors/
which according to readme.md at the same location is CC0.
I created this by typing "sup" into TextEdit.app on macOS 13.4,
hitting Cmd-P to bring up the print dialog, clicked the PDF button
at the bottom, changed Title and Author to "sup", clicked
"Security Options…", and checked "Require password to open document"
(with password "sup").
This file tests several things:
- It has a compressed stream as first object. This used to make the
linearization dict detection logic assert.
- It uses AES as encryption key using version 4 of the encryption
dict. This used to not be implemented.
Max width shouldn't be tied to min width, commit d33b99d went too far
and made them the same when the table-root had a specified percentage
width.
Fixes#19940.
Since the underlying HTML::Window can change, caching property accesses
on WindowProxy is not as simple as remembering the shape. Let's disable
caching here for now. We can come back to it in the future when we have
no low-hanging fruit left. :^)
Fixes an assertion failure on https://twinings.co.uk/
The previous iteration of this API was somewhat odd and rough in random
places, which degraded usability and made less than perfect sense.
This commit reworks the API to be a little closer to more
conventional promise APIs (a la javascript promises).
Also adds a test to ensure the class even works.
