In this bug we're moving away from monolithic JNI headers to class-specific
headers so that we don't have to rebuild the world every time we make a change
to a JNI interface.
Differential Revision: https://phabricator.services.mozilla.com/D75366
We already ensure that we finish our write before doing a fast
shutdown, but right now there's just too much noise in the late
write checking telemetry because of this.
Differential Revision: https://phabricator.services.mozilla.com/D75389
In the tests we now explicitly use "Main Process" (or similar) for the process
name in various places to avoid having to filter out a PID.
Also in `test_aboutmemory[34].xhtml`, the main process entries now appear
first, as they should, because the sorting code now looks for a "Main Process"
prefix, rather than an exact match.
Depends on D72995
Differential Revision: https://phabricator.services.mozilla.com/D73164
Currently the JS holders table is represented as a map which contains pointers to entries in a SegmentedVector. This patch keeps the single map but use a vector per zone and also has a catch-all vector for where we don't know the zone or the holder can have pointers to more than one zone.
Differential Revision: https://phabricator.services.mozilla.com/D68522
Currently we get a rooting hazard when nsWrapperCache::UpdateWrapperForNewGlobal calls ReleaseWrapper because the analysis can't see through the virtual method call.
Differential Revision: https://phabricator.services.mozilla.com/D71600
Currently, to remove a holder entry from a vector we swap the entry with the last one in the vector and then shrink the vector (and fix up the map). With per-zone vectors we won't have a reference to the vector to get the last entry. One solution would be to store the zone in the entry and look up the vector in a map but I'd like to avoid this if possible because of the space overhead and the extra lookup.
This patch defers clears the entry's pointers when it is removed and actually removes it when the vector is next iterated.
Differential Revision: https://phabricator.services.mozilla.com/D68518
Prior to this patch, the startupcache created its own mWriteThread off which it
wrote to disk. It's initialized by MaybeSpawnWriteThread, which got called
at shutdown, to do the shutdown write if there was any reason to do so, and
from a timer that is re-initialized after every addition to the startup cache,
to run 60s after the last change to the cache.
It then joined that write thread on the main thread (in other words, blocks
on that off-main-thread write completing from the main thread) when:
- xpcom-shutdown fired
- the startupcache itself gets destroyed
- someone calls any of:
* HasEntry
* GetBuffer
* PutBuffer
* InvalidateCache
This patch removes the separate write thread, and instead dispatches a task to
the background task queue, indicating it can block. The task is started in
the same circumstances where we previously used to write (timer from the last
PutBuffer call, and shutdown if necessary).
To ensure it cannot be trying to use the data it writes out (mTable) from
the other thread while that data changes on the main thread, we use a mutex.
The task locks the mutex before starting, and unlocks when finished.
Enumerating the cases that we used to block on joining the thread:
In terms of application shutdown, we expect the background task queue to
either finish the write task, or fail to run it if it hasn't started it yet.
In the FastStartup case, we check if a write was necessary; if so, we
attempt to gain the lock without waiting. If we're successful, the write has
not yet started, and we instead run the write on the main thread. Otherwise,
we retry gaining the lock, blocking this time, thus guaranteeing the
off-the-main-thread write completes.
The task keeps a reference to the startupcache object, so it cannot be
destroyed while the task is pending.
Because the write does not modify `mTable`, and neither does `HasEntry`,
we do not need to do anything there.
In the `GetBuffer` case, we do not modify the table unless we have to read
the entry off disk (memmapped into `mCacheData`). This can only happen if
`mCacheData.initialized()` returns true, and we specifically call
`mCacheData.reset()` before firing off the write task to avoid this.
`mCacheData` is only re-initialized if someone calls `LoadArchive()`,
which can only happen from `Init()` (which is guaranteed not to run
again because this is a singleton), or `InvalidateCache()`, where we lock
the mutex (see below). So this is safe - but we assert on the lock to try
and avoid people breaking this chain of assumptions in the future.
When `PutBuffer` is called, we try to lock the mutex - but if locking fails
(ie the background thread is writing), we simply fail to store the entry
in the startupcache. In practice, this should be rare - it'd happen if
new calls to PutBuffer happen while writing during shutdown (when really,
we don't care) or when it's been 60 seconds since the last PutBuffer so
we started writing the startupcache.
When InvalidateCache is called, we lock the mutex - we shouldn't try to
write while invalidating, or invalidate while writing. This may be slow,
but in practice nothing should call `InvalidateCache` except developer
restarts or the `-purgecaches` commandline flag, so it shouldn't
matter a great deal.
Differential Revision: https://phabricator.services.mozilla.com/D70413
--HG--
extra : moz-landing-system : lando
Unlinking, such as in UnbindFromTree, can add runnables that keep alive CCed
objects after they have been unlinked. In combination with some unknown
fields not being traversed and unlinked this could cause intermittent shutdown
leaks, if those unknown fields hold alive objects that need the CC to be
collected. Work around this by clearing out these runnables after every
shutdown CC.
Also, fix the order of these two steps, which was apparently wrong
before.
Differential Revision: https://phabricator.services.mozilla.com/D71049
--HG--
extra : moz-landing-system : lando
This pattern:
```
native nsSize (nsSize);
^
```
Causes a parsing error in `ply` 3.10. This can be easily fixed by removing the space and reformatting to this:
```
native nsSize(nsSize);
```
Differential Revision: https://phabricator.services.mozilla.com/D70711
--HG--
extra : moz-landing-system : lando
L3 cache being present in /proc/cpuinfo is an Intel-ism. Use the cross-platform
/sys/devices instead.
Differential Revision: https://phabricator.services.mozilla.com/D70892
--HG--
extra : moz-landing-system : lando
The patch makes sure the checkSingleZoneHolders variable is always present. I expect it will get optimised away in builds where it's unused.
Differential Revision: https://phabricator.services.mozilla.com/D70818
--HG--
extra : moz-landing-system : lando
The patch makes sure the checkSingleZoneHolders variable is always present. I expect it will get optimised away in builds where it's unused.
Differential Revision: https://phabricator.services.mozilla.com/D70818
--HG--
extra : moz-landing-system : lando
Through their use of Smprintf, the existing warning message formatting
is resistant to OOM errors. So I figured that we should probably use
something that doesn't infallibly allocate like GetErrorName does.
Differential Revision: https://phabricator.services.mozilla.com/D70771
--HG--
extra : moz-landing-system : lando
TabGroup never really made any difference in which thread something go
dispatched to. This was the intended use, but development of TabGroups
with abstract main threads never made it that far. The good thing is
that thish makes it safe to also remove to the SystemGroup and instead
switch all SystemGroup dispatches to dispatches to main thread.
Timers for setTimeout and workers were the sole users of wrapped and
throttled event targets, that those throttled queues have been moved
to the BrowsingContextGroup and are now accessed explicitly.
The SchedulerEventTarget has been removed, since there are no longer a
separate event target for every TaskCategory. Instead a
LabellingEventTarget has been added to DocGroup to handle the case
where an event is dispatched do DocGroup or when an AbstractThread is
created using a DocGroup. This means that we'll actually label more
events correctly with the DocGroup that they belong to.
DocGroups have also been moved to BrowsingContextGroup.
Depends on D67636
Differential Revision: https://phabricator.services.mozilla.com/D65936
--HG--
extra : moz-landing-system : lando
This patch also tries to remove the event target entirely if it would
default to the main thread on a null event target.
Depends on D67634
Differential Revision: https://phabricator.services.mozilla.com/D67635
--HG--
extra : moz-landing-system : lando
To be able to remove SystemGroup, NS_ReleaseOnMainThreadSystemGroup
needs to have its dependency on SystemGroup removed. Since all
releases using SystemGroup would've released on the main thread anyway
we can safely replace NS_ReleaseOnMainThreadSystemGroup with
NS_ReleaseOnMainThread.
Depends on D64390
Differential Revision: https://phabricator.services.mozilla.com/D67631
--HG--
extra : moz-landing-system : lando
Now that we have some per-zone vectors we can skip tracing those for zones that are not being collected.
Differential Revision: https://phabricator.services.mozilla.com/D68523
--HG--
extra : moz-landing-system : lando
Currently the JS holders table is represented as a map which contains pointers to entries in a SegmentedVector. This patch keeps the single map but use a vector per zone and also has a catch-all vector for where we don't know the zone or the holder can have pointers to more than one zone.
Differential Revision: https://phabricator.services.mozilla.com/D68522
--HG--
extra : moz-landing-system : lando
Currently, to remove a holder entry from a vector we swap the entry with the last one in the vector and then shrink the vector (and fix up the map). With per-zone vectors we won't have a reference to the vector to get the last entry. One solution would be to store the zone in the entry and look up the vector in a map but I'd like to avoid this if possible because of the space overhead and the extra lookup.
This patch defers clears the entry's pointers when it is removed and actually removes it when the vector is next iterated.
Differential Revision: https://phabricator.services.mozilla.com/D68518
--HG--
extra : moz-landing-system : lando
Also adds missing includes in some files, these were previously only transivitely
included through mozilla/TypeTraits.h.
Differential Revision: https://phabricator.services.mozilla.com/D68561
--HG--
extra : moz-landing-system : lando
Aaron Klotz
Doug Thayer
Valentin Gosu
Chris Fronk
Chris Peterson
Simon Giesecke
Nicholas Nethercote
Eric Rahm
Bogdan Tara
Jon Coppeard
Peter Van der Beken
Tom Schuster
Sylvestre Ledru
Ciure Andrei
Emilio Cobos Álvarez
Andrew McCreight
Gijs Kruitbosch
Kristen Wright
Chris H-C
Cameron McCormack
Emma Malysz
Gerald Squelart
Andreas Farre
André Bargull