This splits out the inner bit of RecvEmptyTransaction to just iterate over
the documents once, rather than iterating over them individually. Originally
I ran into difficulties with this and then left it on the table, but I think
it was enabled by splitting out the epochs in pipeline info by renderroot.
Differential Revision: https://phabricator.services.mozilla.com/D35123
--HG--
extra : moz-landing-system : lando
This change replaces and removes code in VRManager that was refactored into the
new VRShMem class.
Differential Revision: https://phabricator.services.mozilla.com/D36986
--HG--
extra : moz-landing-system : lando
Currently items are painted with a context that has a transform of
-mLayerBounds.TopLeft(). This means that if TopLeft() changes the commands
will be in the wrong place because the -TopLeft() offset is baked into the
recording.
I don't think we've ever needed to support painting without this transformed
baked in so there were some infrastructure changes that needed to be made to
make this possible. Most of the problems come from the use of
gfxContext::GetClipExtents which expose the bounds of the underlying surface.
The biggest of these was fixed by the CreateClippedDrawTarget rewrite. The rest
should be handled by ensuring that the DrawTarget has bounds that are at least
as big as the union of the individual item bounds. i.e. GetClipExtents should
never intersect with bounds of the item.
This change has a couple of parts:
1. Store mLayerBounds.TopLeft() in the recording so that it will be subtracted
during replay
2. Use mLayerBounds as the Rect of the RecordingDrawTarget
3. Don't include mLayerBounds.TopLeft() in the transform during recording.
4. Adjust the dirty rect by recordingOrigin before we use it as a clip so that
it stays in the right place.
5. In PaintContainerItem the bounds parameter of PushLayer is in device space
so we need to account for the shift in the location of device space in the
DrawTargetRecording.
Differential Revision: https://phabricator.services.mozilla.com/D37513
--HG--
extra : moz-landing-system : lando
This change replaces and removes code in VRManager that was refactored into the
new VRShMem class.
Differential Revision: https://phabricator.services.mozilla.com/D36986
--HG--
extra : moz-landing-system : lando
I suspect we may change things more in the future as we blob's size
just be the visible rect but this is an incremental step in the right
direction.
It also includes some changes to make sure that we always update our
tiles appropriately.
Differential Revision: https://phabricator.services.mozilla.com/D37079
--HG--
extra : moz-landing-system : lando
This also disables the test on Windows 7 because the newly enabled subtests
fail intermittently there. We don't care so much about Windows 7 these days,
and I don't have a local setup to reproduce it, so I didn't investigate the
failure.
Differential Revision: https://phabricator.services.mozilla.com/D37819
--HG--
extra : moz-landing-system : lando
This will let us do the subtraction of the recording origin during
playback instead of during recording.
It will also let us merge recordings that have different origins.
Differential Revision: https://phabricator.services.mozilla.com/D37561
--HG--
extra : moz-landing-system : lando
This removes duplication and makes it a bit safer by using ConvertFromBytes to
do an unaligned read of the indexOffset.
Also, inner classes can't have template methods. Who knew.
Differential Revision: https://phabricator.services.mozilla.com/D37870
--HG--
extra : moz-landing-system : lando
Also ensure we consistently use the original-case family name in FontNameCache entries,
and only lowercase it to a "key" for lookup/insertion into the font list. This avoids
failures in test_font_whitelist.html due to inconsistency in whether family names have
been lowercased.
Differential Revision: https://phabricator.services.mozilla.com/D36112
--HG--
extra : moz-landing-system : lando
This is the main part of the implementation, except that it doesn't handle populating the
local names table (for @font-face src:local() lookups) with Full and PostScript names;
that follows in the next patch.
Differential Revision: https://phabricator.services.mozilla.com/D36110
--HG--
extra : moz-landing-system : lando
I'll need to add a couple of extra fields to the cache records, and realized that the current format
looks really fragile; in theory, it'd break if a font name ever contained a comma or semicolon
(unlikely though that may be). So let's fix it to be a bit more robust before we add to it further.
Using control characters from the C0 range to delimit fields/records, instead of ASCII punctuation,
removes the (tiny) risk of conflict with characters that actually occur in a name, and using
distinct field and record separators means that we can better check that the records we're loading
from the cache actually match the expected format.
(Given that the startup cache gets re-created when the build ID is updated, a change in the format
between versions wasn't going to cause problems for users; their old cache just gets blown away
when upgrading. Still, a little more robustness seems like a good thing.)
Differential Revision: https://phabricator.services.mozilla.com/D36109
--HG--
extra : moz-landing-system : lando
This makes the functional structure a bit cleaner, so that it'll be easier to slip in the alternative
codepath for the shared font-list.
Differential Revision: https://phabricator.services.mozilla.com/D36108
--HG--
extra : moz-landing-system : lando
A bit of cleanup of the existing code, before we start actually implementing new stuff.
Differential Revision: https://phabricator.services.mozilla.com/D36107
--HG--
extra : moz-landing-system : lando
This isn't really specific to the FT2 fontlist, it's a general fixup that I noticed while
working on this. (The missing checks aren't crucial, but mean that we might re-read a cmap
when we shouldn't need to.)
Differential Revision: https://phabricator.services.mozilla.com/D36106
--HG--
extra : moz-landing-system : lando
This implements the idea of automatically setting a content proc's
render root based on the render root enclosing the iframe that
points to it. There was a bit of cleanup in here that was a bit
tricky to extract from the core patch revolving around how we
use the Api(...) helper. This was to avoid the situation where
we use the Api(...) helper before our render root is initialized,
when we don't actually have to. I.e., when we just want the root
WebRenderAPI in all cases.
An alternative to this approach could be to fully built out the
WebRender transactions and just queue those up to be sent. However,
transaction building has various side effects which are committed
before the transaction is actually sent, so we would have to build
out some scheme for deferring those as well. This seemed simpler.
Patch primarily written by :dthayer
Differential Revision: https://phabricator.services.mozilla.com/D37078
--HG--
extra : moz-landing-system : lando
This splits out the inner bit of RecvEmptyTransaction to just iterate over
the documents once, rather than iterating over them individually. Originally
I ran into difficulties with this and then left it on the table, but I think
it was enabled by splitting out the epochs in pipeline info by renderroot.
Differential Revision: https://phabricator.services.mozilla.com/D35123
--HG--
extra : moz-landing-system : lando
This minimizes regret by requiring T: Copy and switches
to read_unaligned() because the pointer can be unaligned.
Differential Revision: https://phabricator.services.mozilla.com/D37861
--HG--
extra : moz-landing-system : lando
Replace `serde`-derived `bincode` with custom binary
serialization/deserialization that generates more efficient code at rustc
`opt-level = 2`.
Differential Revision: https://phabricator.services.mozilla.com/D32782
--HG--
extra : moz-landing-system : lando
This implements the idea of automatically setting a content proc's
render root based on the render root enclosing the iframe that
points to it. There was a bit of cleanup in here that was a bit
tricky to extract from the core patch revolving around how we
use the Api(...) helper. This was to avoid the situation where
we use the Api(...) helper before our render root is initialized,
when we don't actually have to. I.e., when we just want the root
WebRenderAPI in all cases.
An alternative to this approach could be to fully built out the
WebRender transactions and just queue those up to be sent. However,
transaction building has various side effects which are committed
before the transaction is actually sent, so we would have to build
out some scheme for deferring those as well. This seemed simpler.
Patch primarily written by :dthayer
Differential Revision: https://phabricator.services.mozilla.com/D37078
--HG--
extra : moz-landing-system : lando
This splits out the inner bit of RecvEmptyTransaction to just iterate over
the documents once, rather than iterating over them individually. Originally
I ran into difficulties with this and then left it on the table, but I think
it was enabled by splitting out the epochs in pipeline info by renderroot.
Differential Revision: https://phabricator.services.mozilla.com/D35123
--HG--
extra : moz-landing-system : lando
This implements the idea of automatically setting a content proc's
render root based on the render root enclosing the iframe that
points to it. There was a bit of cleanup in here that was a bit
tricky to extract from the core patch revolving around how we
use the Api(...) helper. This was to avoid the situation where
we use the Api(...) helper before our render root is initialized,
when we don't actually have to. I.e., when we just want the root
WebRenderAPI in all cases.
An alternative to this approach could be to fully built out the
WebRender transactions and just queue those up to be sent. However,
transaction building has various side effects which are committed
before the transaction is actually sent, so we would have to build
out some scheme for deferring those as well. This seemed simpler.
Patch primarily written by :dthayer
Differential Revision: https://phabricator.services.mozilla.com/D37078
--HG--
extra : moz-landing-system : lando
This splits out the inner bit of RecvEmptyTransaction to just iterate over
the documents once, rather than iterating over them individually. Originally
I ran into difficulties with this and then left it on the table, but I think
it was enabled by splitting out the epochs in pipeline info by renderroot.
Differential Revision: https://phabricator.services.mozilla.com/D35123
--HG--
extra : moz-landing-system : lando
Replace `serde`-derived `bincode` with custom binary
serialization/deserialization that generates more efficient code at rustc
`opt-level = 2`.
Differential Revision: https://phabricator.services.mozilla.com/D32782
--HG--
extra : moz-landing-system : lando
This change introduces the stubs for nsFxrCommandLineHandler,
which will support launching Firefox Reality on Desktop.
Differential Revision: https://phabricator.services.mozilla.com/D37760
--HG--
extra : moz-landing-system : lando
Replace `serde`-derived `bincode` with custom binary
serialization/deserialization that generates more efficient code at rustc
`opt-level = 2`.
Differential Revision: https://phabricator.services.mozilla.com/D32782
--HG--
extra : moz-landing-system : lando
This will let us do the subtraction of the recording origin during
playback instead of during recording.
It will also let us merge recordings that have different origins.
Differential Revision: https://phabricator.services.mozilla.com/D37561
--HG--
extra : source : cb1d78b00e25dd7fcfec86216c7a83c85ce9a982
extra : histedit_source : 68a31189f05998ba3b8a29a624d7ebe37636c4d9
I suspect we may change things more in the future as we blob's size
just be the visible rect but this is an incremental step in the right
direction.
It also includes some changes to make sure that we always update our
tiles appropriately.
Differential Revision: https://phabricator.services.mozilla.com/D37079
--HG--
extra : moz-landing-system : lando
The APZ tree walk is recursive but the render root was not being updated when
walking up out of a subtree with a different render root. This changes the
code to use a stack and push/pop the render root for subtrees as we enter
and exit the subtrees as part of the tree walk.
Differential Revision: https://phabricator.services.mozilla.com/D37614
--HG--
extra : moz-landing-system : lando
Only the parent process ever respects the renderroot attribute, so we can
add some extra early-exit checks for this. This also adds a bit of safety
for the next patch, to avoid inadvertently exposing renderroot stuff to
web content.
Also, the GetRenderRootForElement function is only ever called by the
GetRenderRootForFrame function, so let's scope it down to avoid doing
unnecessary work.
Differential Revision: https://phabricator.services.mozilla.com/D37504
--HG--
extra : moz-landing-system : lando
386947-1.xul, now has one assertion since we take a different code
path with chrome URL's and XBL files. The assertion is triggered since the
binding is invalid.
Differential Revision: https://phabricator.services.mozilla.com/D34542
--HG--
extra : moz-landing-system : lando
On startup some program binaries are loaded from disk into an
in-memory cache. When we call create_program() we check if the
required program is present in this cache, and if so we call
glProgramBinary(). This is done early on so that the driver can
perform any necessary work in the background.
There may however be binaries in the disk cache that have not yet been
loaded in to memory, in order not to slow down startup. This change
makes it so that we attempt to load missing binaries from disk during
link_program(). The reason we do not do this in create_program() is
because that would result in loading all shaders from disk during
startup, which we want to avoid. Loading these shaders may therefore
take slightly longer than if they'd been loaded at startup, but will
still be much faster than recompiling them from scratch, and startup
will remain quick.
If loading the shaders on startup had previously timed out, then we do
not attempt to load shaders on demand as the disk is probably too slow
for that to be useful.
Depends on D33954
Differential Revision: https://phabricator.services.mozilla.com/D33955
--HG--
extra : moz-landing-system : lando
Webrender caches the program binaries of shaders used within the first
ten frames, so that on next startup it can load them from disk rather
than having to recompile them.
Previously it would load all binaries found in the disk cache on
startup, and when saving to the cache it would delete any existing
binaries that weren't used.
This changes it so that unused binaries are not deleted. The disk
space this requires is insignificant, but as the cache grows loading
all the shaders on startup can get expensive. To solve that we write a
whitelist of the shaders used during startup, and only load those
during the next startup.
Differential Revision: https://phabricator.services.mozilla.com/D33954
--HG--
extra : moz-landing-system : lando
This will let us do the subtraction of the recording origin during
playback instead of during recording.
It will also let us merge recordings that have different origins.
Differential Revision: https://phabricator.services.mozilla.com/D37561
--HG--
extra : moz-landing-system : lando
I suspect we may change things more in the future as we blob's size
just be the visible rect but this is an incremental step in the right
direction.
It also includes some changes to make sure that we always update our
tiles appropriately.
Differential Revision: https://phabricator.services.mozilla.com/D37079
--HG--
extra : moz-landing-system : lando
This lets us record at arbitratry offsets. Without it gfxContext::GetClip()
would break because it uses DrawTarget::GetRect() as the initial rect that it
intersects subsequent clips with. We also can't just use a DrawTargetOffset
because that applies a transform to the inner DrawTarget and will impact the
recorded commands.
Differential Revision: https://phabricator.services.mozilla.com/D37075
--HG--
extra : moz-landing-system : lando
This implements the machinery for the splitting of static prefs headers, and
uses it for a single header. #includes are used in such a way that the amount
of boilerplate for each static prefs header file is minimal.
Future patches will split the remaining prefs into more header files.
Differential Revision: https://phabricator.services.mozilla.com/D36154
--HG--
rename : modules/libpref/StaticPrefs.h => modules/libpref/StaticPrefsBase.h
rename : modules/libpref/StaticPrefs.h => modules/libpref/init/StaticPrefListBegin.h
extra : moz-landing-system : lando
Before this patch, we would use fallback for all border images. Now for
all but vector images we will use the WebRender border images
primitives. Vector images are an exception because the fallback is
clever in that it upscales the vector image and clips to only draw the
region it requires. This avoids artifacting but to do something similar
for WebRender as it is currently defined, we would increase our CPU and
memory footprint as we would need to produce the entire vector image
upscaled, not just the parts we need. In the future we should change
WebRender to accept different image resources for each of the segments.
Differential Revision: https://phabricator.services.mozilla.com/D37093
When double buffering is enable with compositor, DXGI_SWAP_EFFECT_FLIP_SEQUENTIAL is used without DirectComposition. There are some devices that swap chain with DXGI_SWAP_EFFECT_FLIP_SEQUENTIAL does not work well. In this case, device reset happens very often. To avoid the situation, the double buffering needs to be disabled when device reset happens.
Differential Revision: https://phabricator.services.mozilla.com/D36579
--HG--
extra : moz-landing-system : lando
it's very helpful to see the list of clips and the way they affect a chased primitive
Example:
```
building clip chain instance with local rect TypedRect(1561.0×1968.0 at (-300.0,-300.0))
clip Rectangle(3840.0×1874.0, Clip) at (0.0,0.0) in space SpatialNodeIndex(1)
flags (empty), resulted in Partial
clip Rectangle(3840.0×1874.0, Clip) at (0.0,0.0) in space SpatialNodeIndex(2)
flags (empty), resulted in Partial
```
Differential Revision: https://phabricator.services.mozilla.com/D37137
--HG--
extra : moz-landing-system : lando
And with this, all tests pass on tryserver when the shared list is enabled.
Differential Revision: https://phabricator.services.mozilla.com/D36112
--HG--
extra : moz-landing-system : lando
This is the main part of the implementation, except that it doesn't handle populating the
local names table (for @font-face src:local() lookups) with Full and PostScript names;
that follows in the next patch.
Differential Revision: https://phabricator.services.mozilla.com/D36110
--HG--
extra : moz-landing-system : lando
I'll need to add a couple of extra fields to the cache records, and realized that the current format
looks really fragile; in theory, it'd break if a font name ever contained a comma or semicolon
(unlikely though that may be). So let's fix it to be a bit more robust before we add to it further.
Using control characters from the C0 range to delimit fields/records, instead of ASCII punctuation,
removes the (tiny) risk of conflict with characters that actually occur in a name, and using
distinct field and record separators means that we can better check that the records we're loading
from the cache actually match the expected format.
(Given that the startup cache gets re-created when the build ID is updated, a change in the format
between versions wasn't going to cause problems for users; their old cache just gets blown away
when upgrading. Still, a little more robustness seems like a good thing.)
Differential Revision: https://phabricator.services.mozilla.com/D36109
--HG--
extra : moz-landing-system : lando
This makes the functional structure a bit cleaner, so that it'll be easier to slip in the alternative
codepath for the shared font-list.
Differential Revision: https://phabricator.services.mozilla.com/D36108
--HG--
extra : moz-landing-system : lando
A bit of cleanup of the existing code, before we start actually implementing new stuff.
Differential Revision: https://phabricator.services.mozilla.com/D36107
--HG--
extra : moz-landing-system : lando
This isn't really specific to the FT2 fontlist, it's a general fixup that I noticed while
working on this. (The missing checks aren't crucial, but mean that we might re-read a cmap
when we shouldn't need to.)
Differential Revision: https://phabricator.services.mozilla.com/D36106
--HG--
extra : moz-landing-system : lando
The original patch in bug 1422530 checked whether we're currently hiding a given piece of text
because we're waiting for a resource to download, and used this as a signal that we should not
initiate another download.
However, that's not really a good criterion to use when we're doing font selection for a given
character, both because it's an indirect signal -- whether painting the text is suppressed will
depend on timing and the font-display property -- and because it doesn't consider whether any
resource that's already being downloaded will actually be relevant for the specific character
we're trying to display.
So this patch instead checks for a relevant in-progress load earlier in the font list during
the specific FindFontForChar call, so that when unicode-range is used to split a font across
multiple subsets, an in-progress load for one subset won't make us defer the downloads of other
subsets if they are also needed for the characters present in the text.
With this, the repeated blinking no longer happens on the site here, as loading of all the
required font subsets is initiated early.
Differential Revision: https://phabricator.services.mozilla.com/D35634
--HG--
extra : moz-landing-system : lando
This uses positive-but-empty #if conditions with else clauses rather than
inverted conditions for better readability and documentation.
Differential Revision: https://phabricator.services.mozilla.com/D37083
--HG--
extra : moz-landing-system : lando
it's very helpful to see the list of clips and the way they affect a chased primitive
Example:
```
building clip chain instance with local rect TypedRect(1561.0×1968.0 at (-300.0,-300.0))
clip Rectangle(3840.0×1874.0, Clip) at (0.0,0.0) in space SpatialNodeIndex(1)
flags (empty), resulted in Partial
clip Rectangle(3840.0×1874.0, Clip) at (0.0,0.0) in space SpatialNodeIndex(2)
flags (empty), resulted in Partial
```
Differential Revision: https://phabricator.services.mozilla.com/D37137
--HG--
extra : moz-landing-system : lando
a follow-up to D36603 that switches the base space from the surface node to the raster node.
Differential Revision: https://phabricator.services.mozilla.com/D36828
--HG--
extra : moz-landing-system : lando
We should also check IsRootContentDocumentCrossProcess instead of
IsRootContentDocument there, it will be fixed in bug 1562505.
The test case in this commit is almost copied-n-pasted from
helper_scroll_into_view_bug1516056.html.
Differential Revision: https://phabricator.services.mozilla.com/D36556
--HG--
extra : moz-landing-system : lando
Seems like GeckoSurfaceTexture::DestroyUnused can change the current
context.
Differential Revision: https://phabricator.services.mozilla.com/D36685
--HG--
extra : moz-landing-system : lando
This change brings the related ShMem code from VRManager and VRService into a
new class, VRShMem. This is to support future work where this ShMem will be used
for other efforts. Having this code in the same class will enable it to be more
easily shared in these efforts.
Until the new class replaces the code in VRManager and VRService, it can be
exercised and validated with two instances of vrtesthost, with the -testmgr and
-testsvc parameters.
Differential Revision: https://phabricator.services.mozilla.com/D36649
--HG--
extra : moz-landing-system : lando
Now that all static pref getters use snake_case, these renamings make sense:
- Getfoo_bar_bazPrefName() -> GetPrefName_foo_bar_baz()
- Getfoo_bar_bazPrefDefault() -> GetPrefDefault_foo_bar_baz()
Differential Revision: https://phabricator.services.mozilla.com/D36563
--HG--
extra : moz-landing-system : lando
* Add a script for running wrench under various software rasterizers.
* Add support to wrench for non-blocking event loop.
* Add support to wrench for selecting GL/ES rendering API.
* Update x11 bindings for wrench, to fix a release only crash.
Differential Revision: https://phabricator.services.mozilla.com/D36551
--HG--
extra : moz-landing-system : lando
* Add a script for running wrench under various software rasterizers.
* Add support to wrench for non-blocking event loop.
* Add support to wrench for selecting GL/ES rendering API.
* Update x11 bindings for wrench, to fix a release only crash.
Differential Revision: https://phabricator.services.mozilla.com/D36551
--HG--
extra : moz-landing-system : lando
we save the native fonts by their full path now. On macOS, there is no
such thing as a full filesystem path for a CGFont (or at least we don't track it),
so loading a capture falls back to the old logic of using the dummy font.
Differential Revision: https://phabricator.services.mozilla.com/D36604
--HG--
extra : moz-landing-system : lando
Since JSWindowActors don't have direct access to synchronous messaging,
ChromeScript callers are going to need to migrate to asynchronous messaging
and queries instead.
Since there's no comparable API to sendQuery for frame message managers, this
patch adds a stub that uses synchronous messaging, but makes the API appear
asynchronous, and migrates callers to use it instead of direct synchronous
messaging. This will be replaced with a true synchronous API in the actor
migration.
Fortunately, most of the time, this actually leads to simpler code. The
`sendQuery` API doesn't have the odd return value semantics of
`sendSyncMessage`, and can usually just be used as a drop-in replacement. Many
of the `sendSyncMessage` callers don't actually use the result, and can just
be changed to `sendAsyncMessage`. And many of the existing async messaging
users can be changed to just use `sendQuery` rather than sending messages and
adding response listeners.
However, the APZ code is an exception. It relies on intricate properties of
the event loop, and doesn't have an easy way to slot in promise handlers, so I
migrated it to using sync messaging via process message managers instead.
Differential Revision: https://phabricator.services.mozilla.com/D35055
--HG--
extra : rebase_source : d5707e87f293a831a5cf2e0b0a7e977090267f78
extra : source : 75ebd6fce136ab3bd0e591c2b8b2d06d3b5bf923
The SpecialPowers set*Pref/get*Pref APIs currently use synchronous messaging
to set and get preference values from the parent process. Aside from directly
affecting callers of those APIs, it also affects callers of `pushPrefEnv`,
which is meant to be asynchronous, but is in practice usually synchronous due
to the synchronous messaging it uses.
This patch updates the getPref APIs to use the in-process preference service
(which most callers are expecting anyway), and also updates the callers of
the setPref and pushPrefEnv APIs to await the result if they're relying on it
taking effect immediately.
Unfortunately, there are some corner cases in tests that appear to only work
because of the quirks of the current sync messaging approach. The synchronous
setPref APIs, for instance, trigger preference changes in the parent
instantly, but don't update the values in the child until we've returned to
the event loop and had a chance to process the notifications from the parent.
The differnce in timing leads some tests to fail in strange ways, which this
patch works around by just adding timeouts.
There should be follow-ups for test owners to fix the flakiness.
Differential Revision: https://phabricator.services.mozilla.com/D35054
--HG--
extra : rebase_source : 941298157e7c82f420cf50ce057154ce9b85301c
extra : source : 189dc8a359815e059a4a217f788d183260bb2bfe
This is the first piece of the blob-recoord series. Adding
some checks to ensure things stay sane.
Differential Revision: https://phabricator.services.mozilla.com/D35806
--HG--
extra : moz-landing-system : lando
