Currently if you don't specify the '-m' option, a fake do-nothing memory
profiler thread gets created instead of the real one. It ignores all
events except for `Exit`. And the timer thread doesn't get created so no
`Print` events are sent.
This changeset instead always creates the real thread. If you specify
'-m' the *timer* thread is still absent and it won't print anything.
However, the memory profiler thread will respond to register/unregister
events, which is good, because if there's a bug in those (e.g.
double-registration of a particular name) it'll show up in invocations
that lack '-m'.
This changeset implements the beginnings of fine-grained measurement of
Servo's data structures.
- It adds a new `SizeOf` trait, which is used to measure the memory used
by heap data structures, and implements it for some std types: Box,
String, Option, Arc, Vec, and DList.
- It adds a new `MemoryReporter` trait which is used to report memory
measurements from other threads to the memory profiler. Reporters are
registered and unregistered with the memory profiler, and the memory
profiler makes measurement requests of reporters when necessary.
- It plumbs a MemoryProfilerChan through to the layout task so it can
register a memory reporter.
- It implements the `SizeOf` trait for `DisplayList` and associated
types, and adds a memory reporter that uses it.
The display list hits 14.77 MiB when viewing
tests/html/perf-rainbow.html, and 2.51 MiB when viewing the Guardians of
the Galaxy Wikipedia page from servo-static-suite. Example output:
0.29: display-list::http://www.reddit.com/
0.00: display-list::http://static.adzerk.net/reddit/ads.html?sr=-reddit.com,loggedout&bust2#http://www.reddit.com
0.00: display-list::http://www.reddit.com/static/createadframe.html
There are a number of FIXME comments indicating sub-optimal things. This
is a big enough change for now that doing them as follow-ups seems best.
The history is now recorded per frame, but needs to be exposed in a followup PR.
Also fixes a race condition that occurs loading iframes under heavy CPU load.
This ensures that iframes never do a reflow / layout until they have a valid
window size set from their parent frame.
If the root layer has a zero size rectangle, it will never be
painted, so assume the pipeline is idle in that case.
Re-enable two reftests that were disabled due to this issue.
@jdm This initial version has a few outstanding issues that I wanted to invite input on. Specifically:
1. I had some difficulty finding a home for the `StorageType` enum. Structs defined outside of the `script` module don't seem to be able to use the `#[jstraceable]` annotation and the `net` module (where `StorageTask` lives) doesn't have access to `script`. Per Simon Sapin's suggestion, I worked around this temporarily by creating a `TraceableStorageType` stand-in struct that was traceable and which could be translated into a regular `StorageType` when being sent to the `StorageTask`. Unsure of the best way to resolve this hack. Thoughts?
2. Apart from the `Storage` constructor used in `Window::SessionStorage` and the new `Window::LocalStorage`, there's also a method called `Storage::Constructor`. I'm unclear on what (if anything) will actually invoke this, so I'm not sure which variant of `StorageType` to use here. I've temporarily created an `Unknown` variant of `StorageType` as a placeholder.
3. I discovered that the web platform tests directory's localStorage tests. Many of them now pass despite the configured expectation that they fail. However, several do not pass. Is there a good way for me to add debug logging or otherwise get a sense of which assertion failed / what went wrong?
Thanks for your continued help!
Cycles were being created in the flow tree since absolutely positioned
descendants had pointers to their containing blocks. This adds
WeakFlowRef (based on Weak<T>) and makes these backpointers weak
references. This also harmonizes our custom Arc<T>, FlowRef, to be
consistent with the upstream implementation.
Fixes#4915.
Cycles were being created in the flow tree since absolutely positioned
descendants had pointers to their containing blocks. This adds
WeakFlowRef (based on Weak<T>) and makes these backpointers weak
references. This also harmonizes our custom Arc<T>, FlowRef, to be
consistent with the upstream implementation.
Fixes#4915.
`./mach test tests/wpt/web-platform-tests/html/browsers/history/the-location-interface/security_location_0.sub.htm` is still failing with the same message as reported in #3219.
CGDataProviderCreateWithData just wraps the underlying buffer. The
underlying buffer needs to be kept around until the data provider is
freed. Adding the buffer to the FontTemplateData struct ensures it
sticks around.
`./mach test-wpt` will fail in non-obvious ways unless all wpt submodules have recursively been checked out first. This ensure they have been when running the command in a checkout of Servo that hasn't been bootstrapped yet.
CGDataProviderCreateWithData just wraps the underlying buffer. The
underlying buffer needs to be kept around until the data provider is
freed. Adding the buffer to the FontTemplateData struct ensures it
sticks around.
SSL is broken-ish (eg tw.yahoo.com, html.spec.whatwg.org don't work since we don't verify SAN properly), this flag can let devs bypass the protection for testing purposes.
Looking for some review, due to a couple of issues:
* The vmin/vmax tests fail due to exit status being 1 instead of 0. However, when I manually check them via `./mach run tests/ref/viewport_percentage_vmax_vmin{,a,b,c}.html`, the pages render correctly
* Resizing is incredibly flaky. I've checked (via logging), and the actual calculation of the computed length is correct for a given viewport size. However, the box dimensions end up all over the place. I've attached a screenshot to demonstrate (the box is supposed to cover the top-left quarter of the window).
Fixes#5165.
When a viewport is resized, the computed values for a style containing viewport percentage length units become stale. However, there's no way for those styles to be invalidated after a resize. As a solution, this commit invalidates the computed values cache after a resize has occurred, which is probably over-kill.
A better solution would probably be to track under what conditions computed values remain valid, and invalidate them as indicated.
