The failure here is a <script> that document.writes a whole HTML document, which
happens to have the Gecko style back-end.
Of all the elements written there, there's an <img> element which triggers a
load, which finishes and changes the state of the <img>, posting a restyle event
to its Gecko-backed restyle manager.
At that point, but before the refresh driver ticks, the element is unbound from
the tree, keeping the restyle root bits from the Gecko back-end.
Afterwards, the element is bound to a Servo-backed document, so
ClearRestyleFlagsIfGecko does nothing (since the document pointer is already
set, and points to a Servo-backed document).
Furthermore, it's inserted on a display: none subtree, so we never reach it
during the traversal to remove the flags.
Other alternative fixes for this same issue would be:
* Clearing the Gecko restyle flags in UnbindFromTree. Gecko doesn't seem to
prevent restyle roots from being outside of the document (it just does
nothing when finding them on ProcessPendingRestyles), so it doesn't seem a
very reliable solution.
* Reorder the flag clearing to be before setting the document in BindToTree.
Seems somewhat risky, and for no good reason.
MozReview-Commit-ID: B6iZKhGHc3A
Another plumbing-only change, which is a precursor for the next patch.
ComputedValues are a Servo-only thing, so in order to use nsStyleSizes for both
Gecko and Servo, the ComputedValues sizes must be moved out.
MozReview-Commit-ID: BOnQSzzV0vC
This is a purely non-functional plumbing change. Instead of passing a
SizeOfState and an nsStyleSizes a bunch of places, we pass an nsWindowSizes,
which contains both of them.
This is a necessary precursor for the next patch.
MozReview-Commit-ID: Ek03wDM50rB
There's only one case of sync frame construction from ContentRemoved now, and
it's not on the element being removed, but on the whitespace siblings if needed,
and _only_ when they don't support lazy frame construction.
Basically, this switches all the RecreateFramesForContent calls to use
`aAsyncInsert` (which I changed to an enum class for readability), except when
we're already reframing.
Also, it switches ReframeTextIfNeeded to opt-in into lazy frame construction,
since it's used only when aFlags == CONTENT_REMOVED.
This allows to simplify the DestroyFramesFor API (which I'm happy to rename to
something more meaningful, since now it's something like
DestroyFramesForAndRecreateThemAsync), and do some other consistency cleanups.
A bunch of the ContentRemoved callsites were pretty random at passing
aAsyncInsert, and that was some kind of a mess. This patch ensures consistency,
and makes it impossible to do O(n^2) work when removing DOM nodes, which is
nice.
The underlying reason for this is explained in the description of bug 1377848,
and basically allows us to remove a bunch of Servo hacks on the longer term (a
few of them are going away already, yay!).
MozReview-Commit-ID: 2DrUTxGV8RX
For the Obama wikipedia page, this covers about 85% of the unmeasured
ComputedValues structs. The about:memory output looks like this:
> +---2,443,648 B (02.41%) -- computed-values
> | +--1,088,272 B (01.07%) -- dom
> | +----945,744 B (00.93%) -- non-dom
> | +----409,632 B (00.40%) -- visited
I'm not sure why some CVs are still being missed.
MozReview-Commit-ID: 1bYWwSi4ihn
This patch moves measurement of ComputedValues objects from Rust to C++.
Measurement now happens (a) via DOM elements and (b) remaining elements via
the frame tree. Likewise for the style structs hanging off ComputedValues
objects.
Here is an example of the output.
> ├──27,600,448 B (26.49%) -- active/window(https://en.wikipedia.org/wiki/Barack_Obama)
> │ ├──12,772,544 B (12.26%) -- layout
> │ │ ├───4,483,744 B (04.30%) -- frames
> │ │ │ ├──1,653,552 B (01.59%) ── nsInlineFrame
> │ │ │ ├──1,415,760 B (01.36%) ── nsTextFrame
> │ │ │ ├────431,376 B (00.41%) ── nsBlockFrame
> │ │ │ ├────340,560 B (00.33%) ── nsHTMLScrollFrame
> │ │ │ ├────302,544 B (00.29%) ── nsContinuingTextFrame
> │ │ │ ├────156,408 B (00.15%) ── nsBulletFrame
> │ │ │ ├─────73,024 B (00.07%) ── nsPlaceholderFrame
> │ │ │ ├─────27,656 B (00.03%) ── sundries
> │ │ │ ├─────23,520 B (00.02%) ── nsTableCellFrame
> │ │ │ ├─────16,704 B (00.02%) ── nsImageFrame
> │ │ │ ├─────15,488 B (00.01%) ── nsTableRowFrame
> │ │ │ ├─────13,776 B (00.01%) ── nsTableColFrame
> │ │ │ └─────13,376 B (00.01%) ── nsTableFrame
> │ │ ├───3,412,192 B (03.28%) -- servo-style-structs
> │ │ │ ├──1,288,224 B (01.24%) ── Display
> │ │ │ ├────742,400 B (00.71%) ── Position
> │ │ │ ├────308,736 B (00.30%) ── Font
> │ │ │ ├────226,512 B (00.22%) ── Background
> │ │ │ ├────218,304 B (00.21%) ── TextReset
> │ │ │ ├────214,896 B (00.21%) ── Text
> │ │ │ ├────130,560 B (00.13%) ── Border
> │ │ │ ├─────81,408 B (00.08%) ── UIReset
> │ │ │ ├─────61,440 B (00.06%) ── Padding
> │ │ │ ├─────38,176 B (00.04%) ── UserInterface
> │ │ │ ├─────29,232 B (00.03%) ── Margin
> │ │ │ ├─────21,824 B (00.02%) ── sundries
> │ │ │ ├─────20,080 B (00.02%) ── Color
> │ │ │ ├─────20,080 B (00.02%) ── Column
> │ │ │ └─────10,320 B (00.01%) ── Effects
> │ │ ├───2,227,680 B (02.14%) -- computed-values
> │ │ │ ├──1,182,928 B (01.14%) ── non-dom
> │ │ │ └──1,044,752 B (01.00%) ── dom
> │ │ ├───1,500,016 B (01.44%) ── text-runs
> │ │ ├─────492,640 B (00.47%) ── line-boxes
> │ │ ├─────326,688 B (00.31%) ── frame-properties
> │ │ ├─────301,760 B (00.29%) ── pres-shell
> │ │ ├──────27,648 B (00.03%) ── pres-contexts
> │ │ └─────────176 B (00.00%) ── style-sets
The 'servo-style-structs' and 'computed-values' sub-trees are new. (Prior to
this patch, ComputedValues under DOM elements were tallied under the the
'dom/element-nodes' sub-tree, and ComputedValues not under DOM element were
ignored.) 'servo-style-structs/sundries' aggregates all the style structs that
are smaller than 8 KiB.
Other notable things done by the patch are as follows.
- It significantly changes the signatures of the methods measuring nsINode and
its subclasses, in order to handle the tallying of style structs separately
from element-nodes. Likewise for nsIFrame.
- It renames the 'layout/style-structs' sub-tree as
'layout/gecko-style-structs', to clearly distinguish it from the new
'layout/servo-style-structs' sub-tree.
- It adds some FFI functions to access various Rust-side data structures from
C++ code.
- There is a nasty hack used twice to measure Arcs, by stepping backwards from
an interior pointer to a base pointer. It works, but I want to replace it
with something better eventually. The "XXX WARNING" comments have details.
- It makes DMD print a line to the console if it sees a pointer it doesn't
recognise. This is useful for detecting when we are measuring an interior
pointer instead of a base pointer, which is bad but easy to do when Arcs are
involved.
- It removes the Rust code for measuring CVs, because it's now all done on the
C++ side.
MozReview-Commit-ID: BKebACLKtCi
We're going to start tracking the restyle root on the presshell, so we'll need
one. This should be fine, since if the presshell doesn't exist yet we can't
have done the initial style, and if it's already been destroyed we don't need
restyle state anymore.
MozReview-Commit-ID: EfNVloI9ENQ
This function is large enough that it doesn't really make sense to have inline,
and we'll be adding more to it in the coming patches.
MozReview-Commit-ID: AnDfzwsMvNy
This fixes the testcase in the bug, which removes and reinserts
some elements. Our invariants require us not to set the dirty
descendants bits on unstyled elements.
MozReview-Commit-ID: 1eESZjNSURG
The patch provides FFI access to Gecko's SeenPtrs type from Rust, in order to
record what has already been measured when measuring Arcs. (The SeenPtrs must
be initialized on the Gecko side because the same table is reused for measuring
all Elements within a window, because Elements can share ComputedValues.) I
have confirmed with DMD that this is working correctly.
The patch also introduces MallocSizeOfRepeats, which is like MallocSizeOf but
takes a SizeOfState, which holds a SeenPtrs table.
MozReview-Commit-ID: DHS8zvCsEdQ
I've been having problems with interdiffs on mozreview lately, so for
ease of review, this patch is being submitted as a seperate patch for
review. Once it is r+'d, it will be folded into the first patch in
this set before landing.
MozReview-Commit-ID: CS9MngaXlBd
All the SizeOf{In,Ex}cludingThis() functions take a MallocSizeOf function
which measures memory blocks. This patch introduces a new type, SizeOfState,
which includes a MallocSizeOf function *and* a table of already-measured
pointers, called SeenPtrs. This gives us a general mechanism to measure
graph-like data structures, by recording which nodes have already been
measured. (This approach is used in a number of existing reporters, but not in
a uniform fashion.)
The patch also converts the window memory reporting to use SizeOfState in a lot
of places, all the way through to the measurement of Elements. This is a
precursor for bug 1383977 which will measure Stylo elements, which involve
Arcs.
The patch also converts the existing mAlreadyMeasuredOrphanTrees table in the
OrphanReporter to use the new mechanism.
Unfortunately, nsGenericHTMLElement::GetAssociatedEditor() cannot use concrete classes because it may return nsIEditor which is set via nsIDocShell.editor. The editor set to nsIDocShell may be implemented by JS since nsIEditor isn't marked as builtinclass.
MozReview-Commit-ID: 6GY9LOYp4hM
It logically makes more sense for these effects to happen after the attribute has actually been changed and moving them allows us to get rid of the member variable HTMLImageElement::mForceReload.
MozReview-Commit-ID: IJBF3AHVb0U
This is necessary to facilitate the transition to cloning attributes instead of reparsing them.
HTMLImageElement's side effects proved to be a bit trickier than those of many other classes because HTMLImageElement::SetAttr intentionally forces an image reload, even if the attribute value has not been changed. Element::SetAttr, on the other hand, usually ignores attribute changes that do not change the attribute value, exiting before BeforeSetAttr is even called. In order to preserve this behavior, another virtual function |OnAttrSetButNotChanged| was added to the Element class. This function will be called in the case that Element::SetAttr exits early, allowing a forced reload to take place at that time.
MozReview-Commit-ID: 4CrH30bo5GT
In order to facilitate the movement of code with side-effects called by Element::SetAttr to Element::BeforeSetAttr and Element::AfterSetAttr, Element::AfterSetAttr should have access to the old value of the attribute. This includes information about whether there was previously a value set or not.
Accomplishing this involved passing an additional argument through functions that find and change the old attribute value in order to ensure that we can differentiate between an empty old value and an absent old value (attribute was not set).
Note that while I tried to ensure that accurate values (and their absence) are reported to Element::AfterSetAttr, I largely ignored SVG. While the old value reported for SVG values should be however accurate the value already being reported to SetAttrAndNotify was, SVG elements do not currently report unset values properly because they will never pass a null pointer to SetAttrAndNotify.
MozReview-Commit-ID: K1mha8CNFZP
