This allows supporting image-set(), etc, and simplifies the bullet frame
code significantly, too thanks to two changes:
* Instead of manually managing the image request, use the CSS image
loader, with the `REQUEST_REQUIRES_REFLOW` flag, to handle image
loads correctly. This didn't exist when this code was initially
implemented, but we can nicely use it now.
* Instead of re-implementing another WebRender command-builder thing,
we can just reuse the nsImageRenderer code.
Differential Revision: https://phabricator.services.mozilla.com/D100774
This allows supporting image-set(), etc, and simplifies the bullet frame
code significantly, too thanks to two changes:
* Instead of manually managing the image request, use the CSS image
loader, with the `REQUEST_REQUIRES_REFLOW` flag, to handle image
loads correctly. This didn't exist when this code was initially
implemented, but we can nicely use it now.
* Instead of re-implementing another WebRender command-builder thing,
we can just reuse the nsImageRenderer code.
Differential Revision: https://phabricator.services.mozilla.com/D100774
Making them non-overflow containers is valid but then we can't just
append them to mFrames since that'll put them on the last line which
may result in some lines in-between the pulled child and its
continuation. That violates the invariant that in-flow continuations
in the same parent must be on consecutive lines (unless they are
overflow containers in which case they need to be on the
[Excess]OverflowContainers lists).
(The crash happens because if we push the lines in-between the two
continuations to the OverflowLines, then we won't find the 2nd
continuation on the first line of OverflowLines.)
(Note also that this change triggers the assertion in
DrainExcessOverflowContainersList, and depends on bug 1680406
to fix that and merge the lists.)
Differential Revision: https://phabricator.services.mozilla.com/D99560
Making them non-overflow containers is valid but then we can't just
append them to mFrames since that'll put them on the last line which
may result in some lines in-between the pulled child and its
continuation. That violates the invariant that in-flow continuations
in the same parent must be on consecutive lines (unless they are
overflow containers in which case they need to be on the
[Excess]OverflowContainers lists).
(The crash happens because if we push the lines in-between the two
continuations to the OverflowLines, then we won't find the 2nd
continuation on the first line of OverflowLines.)
(Note also that this change triggers the assertion in
DrainExcessOverflowContainersList, and depends on bug 1680406
to fix that and merge the lists.)
Differential Revision: https://phabricator.services.mozilla.com/D99560
Precomputing the skipBEnd bit is odd / wrong. Using the PreReflow
version causes no regression, and allows us to simplify the code.
It also reverts the test annotations added to bug 1675376 which were
caused by the extra argument to GetLogicalSkipSides() somehow.
Differential Revision: https://phabricator.services.mozilla.com/D97418
Precomputing the skipBEnd bit is odd / wrong. Using the PreReflow
version causes no regression, and allows us to simplify the code.
It also reverts the test annotations added to bug 1675376 which were
caused by the extra argument to GetLogicalSkipSides() somehow.
Differential Revision: https://phabricator.services.mozilla.com/D97418
That avoids going all the way to the first continuation to call ResolveBidi on
it. This shaves a bunch of time when there are a lot of pages.
This is more problematic than it seems specially when there's no bidi, because
in the "bidi not enabled" case we never remove the flag, which is bad. When
bidi is actually enabled we usually have done the resolution already.
Differential Revision: https://phabricator.services.mozilla.com/D97358
This removes virtually all the time under ConsumedBSize. See the comment for
what ensures the correctness of the cache: Basically, we refresh the cache for
a frame continuation every time we reflow it, which means that when next
continuations go look for it it should be up-to-date (we rely on that already
because we're looking at the content rect).
Differential Revision: https://phabricator.services.mozilla.com/D97357
That avoids going all the way to the first continuation to call ResolveBidi on
it. This shaves a bunch of time when there are a lot of pages.
This is more problematic than it seems specially when there's no bidi, because
in the "bidi not enabled" case we never remove the flag, which is bad. When
bidi is actually enabled we usually have done the resolution already.
Differential Revision: https://phabricator.services.mozilla.com/D97358
This removes virtually all the time under ConsumedBSize. See the comment for
what ensures the correctness of the cache: Basically, we refresh the cache for
a frame continuation every time we reflow it, which means that when next
continuations go look for it it should be up-to-date (we rely on that already
because we're looking at the content rect).
Differential Revision: https://phabricator.services.mozilla.com/D97357
This patch is generated via the rename functionality in my editor; add
`mozilla::` prefix to `OverflowAreas` in headers; and remove the
`OverflowType` alias added in Part 1.
Differential Revision: https://phabricator.services.mozilla.com/D97235
It's straightforward to determine which writing mode is required to pass
to ComputedLogicalMargin() by looking at any subsequent method called on
the margin.
For example, if we see
```
ComputedLogicalMargin().BStartEnd(wm);
```
the writing mode needed to pass to `ComputedLogicalMargin()` is `wm`.
Differential Revision: https://phabricator.services.mozilla.com/D95661
Both are aliases to IntrinsicISizeType::MinISize and
IntrinsicISizeType::PrefISize.
Remove MOZ_ASSERT in nsLayoutUtils::IntrinsicForAxis and
nsContainerFrame::DoInlineIntrinsicISize since IntrinsicISizeType is a
enum class nowadays, which cannot have other values.
I've compiled this patch with DEBUG_INTRINSIC_WIDTH defined in
nsLayoutUtils.cpp, and fixed aWM undefined in
nsLayoutUtils::MinSizeContributionForAxis().
Differential Revision: https://phabricator.services.mozilla.com/D94618
`IS_TRUE_OVERFLOW_CONTAINER` is defined in nsContainerFrame.h, but is
already used in several places in nsIFrame. It really should be an
nsIFrame's method.
Differential Revision: https://phabricator.services.mozilla.com/D92007
The three implementations of StealFrame() -- nsContainerFrame,
nsBlockFrame, and nsInlineFrame -- all have assertions to guarantee
aChild can be found, so the operation shouldn't fail.
This change shouldn't change behavior at all.
Differential Revision: https://phabricator.services.mozilla.com/D91996
But discard it when backplating behind text, so that text is readable.
This should be uncontroversial... Dealing with widgets is a bit harder
so TBD.
Differential Revision: https://phabricator.services.mozilla.com/D91779
But discard it when backplating behind text, so that text is readable.
This should be uncontroversial... Dealing with widgets is a bit harder
so TBD.
Differential Revision: https://phabricator.services.mozilla.com/D91779
Currently, to set ComputeSizeFlags, the caller uses an anonymous enum as
a parameter to ReflowInput constructor to set ReflowInputFlags fields,
and then ReflowInput creates a ComputeSizeFlags from the relevant
ReflowInputFlags fields in Init().
This patch simplifies this flags handover by adding ComputeSizeFlags
parameter to ReflowInput so that the caller can create the flags and
pass it to ReflowInput directly. The can also simplifies the process
needed to add a new ComputeSizeFlag.
We still need to store ComputeSizeFlags in ReflowInput since there's one
caller in `nsBlockFrame::ComputeFinalSize` wanting to checking
`ComputeSizeFlag::BClampMarginBoxMinSize`.
Note 1: ComputeSizeFlags is added only to the ReflowInput's constructor
that also takes parent ReflowInput. The other constructor's existing
callers don't need it.
Note 2: I don't bother adjust the value of DUMMY_PARENT_REFLOW_INPUT,
CALLER_WILL_INIT, and STATIC_POS_IS_CB_ORIGIN because they are going to
be converted into a enum class in a later patch.
This change shouldn't change behavior.
Differential Revision: https://phabricator.services.mozilla.com/D89543
We already use SetProperty() extensively for nsFrameList properties like
OutsideMarkerProperty(), BackdropProperty(), etc, so we can simplify the
interface by removing SetPropTableFrames().
Differential Revision: https://phabricator.services.mozilla.com/D88459
Not all APIs added in this patch are used immediately, but for the sake of
completeness, they are all added.
Their document will be updated in Part 3 after adding the relevant APIs for
ExcessOverflowContainersProperty().
nsContainerFrame::DrainExcessOverflowContainersList() has several calls to set
OverflowContainersProperty() whose life cycle need special attention. We will
deal with them later in Part 4.
Differential Revision: https://phabricator.services.mozilla.com/D88456
When a clearance frame is discovered in a BFC
subtree (`ReflowInput::WillReflowAgainForClearance()` is true), all the
block descendants of the BFC return from `ReflowBlockFrame` immediately
without placing their children or requesting continuation for their
children, etc., because the BFC is going to reflow the subtree a
second time.
When this happens, any block descendant whose lines contain pushed
floats should be marked dirty and reflowed again so that they can go
through all the split floats logic and set their nsReflowStatus to
incomplete/overflow-incomplete correctly. (We may develop a smarter way
in bug 851629 to make the performance better.)
The condition `aState.mPresContext->IsPaginated()` is added to prevent
1406291-1.html from failing. In 1406291-1.html, the table containing the
float in its subtree is being pushed to the second page due to force
break, and the float is in PushedFloatsList of its parent block. When we
reflow the block again (in the second page), it is reflowed under the
measuring stage of the table layout with unconstrained available
block-size. nsBlockFrame::DrainSelfPushedFloats() deliberately doesn't
drain pushed floats if their placehoder are in the same block, and
expect the floats to be reflowed again. Before this patch, the line can
be marked dirty via HasPushedFloats(), but in this patch,
HasPushedFloats() is moved into the nested if-else, so we need
`aState.mPresContext->IsPaginated()` to let the block know that it's in
a paginated context, and needs to reflow any line that contains pushed
floats.
Differential Revision: https://phabricator.services.mozilla.com/D85480
Before this patch, a block frame considers a line dirty only if the line's
block-end is larger than content area's block-end.
However, scrollable overflow area should be broken in paginated environment
(both in a column container and in printing). We should take this into account
when marking a line dirty.
The dynamic wpt reftests are written with the following conditions in mind to
expose the bug:
1. Initially, all the elements fit into the first column.
2. After shrinking the multicol height, the block itself still fits into
the first column, but its tall child (either a simple tall block or a
tall float), which overflows the smaller multicol height, needs to be
broken.
The flexbox printing reftest are similar. A flex container uses
unconstrained available block-size to measure flex items' block-size,
and then performs a final reflow with constrained available block-size,
which is similar to the dynamic scenario above. One caveat is that flex
items usually have IsBResize() set. So to expose this bug, a float
element needs to be deeply nested under a flex item.
Differential Revision: https://phabricator.services.mozilla.com/D85479
