There's no good reason why these should't just be constants. The patch also
appends "MiB" to some of the C++ values to make their meaning clearer.
This patch fixes one outright bug, and one inconsistency.
The bug is due to a prefname mismatch:
- ContentPrefs.cpp and AvailableMemoryTracker.cpp use
"memory.low_virtual_mem_threshold_mb".
- all.js uses "memory.low_virtual_memory_threshold_mb".
Which means that "memory.low_virtual_memory_threshold_mb" showed up in
about:config, but if you changed it nothing would happen because the callback
listened for changes to to "memory.low_virtual_mem_threshold_mb"!
Now for the inconsistency. The above means we actually use a value of 256 for
the virtual memory threshold, even though all.js says 128. But we *do* use a
value of 128 for the commit space threshold, because that's what all.js says
and that prefname is used correctly everywhere. The patch changes the commit
space threshold to 256 for consistency with the virtual memory threshold.
What a mess!
--HG--
extra : rebase_source : d3842c732efa9ab0e90eeb6a4f341aeb289589ed
This was originally added for b2g, where the pref had a different value.
Bug 1398033 enabled it everywhere.
--HG--
extra : rebase_source : c8bb03190cd00d25e6c2ec62a99ed8e911e08197
Use MOZ_CRASH, MOZ_CRASH_UNSAFE_OOL, or MOZ_CRASH_UNSAFE_PRINTF instead.
MozReview-Commit-ID: 1kCCHMlgbGP
--HG--
extra : rebase_source : 2f07ced16bccebf30cd3b2b5fea35e9868d32dad
extra : source : 0bf2c8425b828e71de55dd175fd0dad635b4e67d
There should be no functional change here, but we rely upon the new
structure in the next patch in the series. This separates out the
notions of removing a request from the load group (which is always
final, and must be executed outside of synchronous calls from the owner
of the imgRequestProxy) and wanting to readd a request to the load group
as a background request (for multipart images).
The most important addition is mForceDispatchLoadGroup which if true
when imgRequestProxy::RemoveFromLoadGroup is called, will dispatch the
removal from the load group instead of executing it inline. This ensures
safety for any callers (e.g. to CancelAndForgetObserver) as above.
This was created for B2G and isn't really useful otherwise. It only works on
Linux, and it's behind the memory.system_memory_reporter pref, which is false
by default.
The patch also removes LinuxUtils.{h,cpp}, which is no longer used.
--HG--
extra : rebase_source : b97a018be11a79f83855a73b88020bfa86e60f78
And remove unreachable code after MOZ_CRASH_UNSAFE_OOL().
MOZ_CRASH_UNSAFE_OOL causes data collection because crash strings are annotated to crash-stats and are publicly visible. Firefox data stewards must do data review on usages of this macro. However, all the crash strings this patch collects with MOZ_CRASH_UNSAFE_OOL are already collected with NS_RUNTIMEABORT.
MozReview-Commit-ID: IHmJfuxXSqw
--HG--
extra : rebase_source : 031f30934b58a7b87f960e57179641d44aefe5c5
extra : source : fe9f638a56a53c8721eecc4273dcc074c988546e
This lets us replace moz_xstrdup() of string literals with AssignLiteral(),
among other improvements.
--HG--
extra : rebase_source : 9994d8ccb4f196cf63564b0dac2ae6c4370defb4
This change requires introducing nsID::Clone(). Because it's infallible, the
patch also removes some redundant failure-handling code. (nsMemory::Clone() is
also infallible, so this code was redundant even before this change.)
--HG--
extra : rebase_source : ef22757d3fa814320490bf7e19e822b8f0c4bdc3
Bug 1134923 removed the use of those functions in gecko, and left some
for the XPCOM standalone glue. The XPCOM standalone glue was severely
stripped down in bug 1306327, with the effect of removing the
implementation for those functions.
The remains in nsXPCOM.h are:
XPCOM_API(void*) NS_Alloc(size_t aSize);
XPCOM_API(void*) NS_Realloc(void* aPtr, size_t aSize);
XPCOM_API(void) NS_Free(void* aPtr);
With no implementation left, the first arm is never actually used, and
the second arm means every remaining use of those functions in the tree
is a macro expansion to one of moz_xmalloc, moz_xrealloc or free.
--HG--
extra : rebase_source : fd1669abc5a25d8edbd5c3a8522e22a5c3f558e2
Currently the Gecko Profiler defines a moderate amount of stuff when
MOZ_GECKO_PROFILER is undefined. It also #includes various headers, including
JS ones. This is making it difficult to separate Gecko's media stack for
inclusion in Servo.
This patch greatly simplifies how things are exposed. The starting point is:
- GeckoProfiler.h can be #included unconditionally;
- everything else from the profiler must be guarded by MOZ_GECKO_PROFILER.
In practice this introduces way too many #ifdefs, so the patch loosens it by
adding no-op macros for a number of the most common operations.
The net result is that #ifdefs and macros are used a bit more, but almost
nothing is exposed in non-MOZ_GECKO_PROFILER builds (including
ProfilerMarkerPayload.h and GeckoProfiler.h), and understanding what is exposed
is much simpler than before.
Note also that in BHR, ThreadStackHelper is now entirely absent in
non-MOZ_GECKO_PROFILER builds.
It seemingly hasn't been needed since Mac OS 10.7. A diagnostic assertion that
has been in place for a while hasn't caught any uses of it.
--HG--
extra : rebase_source : 9834849eec9174267c7df8de7fd22840ffa36d8f
Add macro for nsISupports inheriting classes that need cycle
collection and only inherit from one parent, with no interface table
needed. Takes care of map and addref/release calls.
MozReview-Commit-ID: LtTwYH0DeEz
--HG--
extra : rebase_source : d3650209ddd7e28cbf205ac84f152b9763c26bde
There is no use mozilla namespace statement in this compilation unit, so the
call to GetErrorName() has to be prefixed with the mozilla namespace.
MozReview-Commit-ID: G3caP2v9PFF
--HG--
extra : rebase_source : 9aebe0711cf8ad81dc47592521a4f29b4f75caab
nsIHandleReportCallback, typedef of nsIMemoryReporterCallback, has been the
preferred name for a long time and is used in most places. This patch removes
nsIMemoryReporterCallback.
--HG--
extra : rebase_source : c675076b4f98d93d96235dad890e31e0b0e6c277
The compiler complains because a postfix increment is done inside of an assert.
MozReview-Commit-ID: LSRsWMn9Tlj
--HG--
extra : rebase_source : 18539a430c3d84675c602699be9a5a62dd477fec
We're calling GetCurrentVirtualThread() which is defined in nsThreadUtils.h, so
without including nsThreadUtils.h we'll get a compile error if unified builds
are disabled.
MozReview-Commit-ID: JWCigMqNJDi
--HG--
extra : rebase_source : bac48b8f91e6561dfb7afe2828c3652431347c40
Free dirty pages appear as "page-cache" in about:memory reports, but
when minimizing memory, and more generally, on memory pressure, they're
currently not force-free()ed on desktop, while they are on mobile.
There doesn't seem to be much reason not to do it on desktop as well,
and it should help reduce the noise in "explicit" allocations measurements
on CI, too.
--HG--
extra : rebase_source : e477532dda5e72eb8797b5bcdc9060ca46106275
ISO C++ forbids converting a string constant to 'wchar_t*' [-Werror=write-strings]
Either change it to a nullptr (which has same intent) or pass through a static
MozReview-Commit-ID: CSunOCyO9PN
--HG--
extra : rebase_source : bfdabc1f463eca75987c6561f7c3ea60acf0340f
These methods return an addrefed raw pointer, which makes them easy to use in
ways that cause leaks. If they're to continue returning an addrefed pointer,
they should explicitly return an already_AddRefed.
This also switches to StaticRefPtr with ClearOnShutdown for the cached
pointers for the sake of sanity.
MozReview-Commit-ID: D0lDpU8Hqug
--HG--
extra : rebase_source : 7b199070805fc0472eaf8409932517700ed23d49
NS_GetWeakReference, called from do_GetWeakReference, QI's its argument
to nsISupportsWeakReference to determine whether a weak reference can be
obtained. If NS_GetWeakReference is already receiving a
nsISupportsWeakReference pointer, or something than can be converted to
one, then we can skip the QI for a small performance win.
QIing to CC interfaces shows up in Speedometer profiles for a few
classes. Presumably there are many of these objects being created and
destroyed. By making these classes check first for the CC interfaces
directly, rather than going up the inheritance chain, this overhead
should be reduced.
MozReview-Commit-ID: I3sf3my8oua
--HG--
extra : rebase_source : f08884a944d5b4ed1eb1da1070de64f21fc9868a
The main purpose of defining this is to make conversion of places that
use the non-CC variant easier. There are many more places that could
be converted to use these new macros, if somebody felt motivated.
MozReview-Commit-ID: HspjcN76fjg
--HG--
extra : rebase_source : bf3baa586f90f0afbe9229c32d38cb34cc909b9b
We now have jemalloc_ptr_info() and moz_malloc_enclosing_size_of(), which can
be used to measure heap blocks via interior pointers. This patch does the
following.
- Adds MOZ_DEFINE_MALLOC_ENCLOSING_SIZE_OF, for defining
measure-via-interior-pointer functions.
- Uses these functions to replace some horrid pointer arithmetic in functions
measuring Rust types.
--HG--
extra : rebase_source : 5128408256c128222025153ae3e0f924b2499a2a
Also adds a mozilla/ResultExtensions.h header to define the appropriate
conversion functions for nsresult and PRResult. This is in a separate header
since those types are not available in Spidermonkey, and this is the pattern
other *Extensions.h headers follow.
Also removes equivalent NS_TRY macros and WrapNSResult inlines that served the
same purpose in existing code, and are no longer necessary.
MozReview-Commit-ID: A85PCAeyWhx
--HG--
extra : rebase_source : a5988ff770888f901dd0798e7717bcf6254460cd
This allows MOZ_TRY and MOZ_TRY_VAR to be transparently used in XPCOM methods
when compatible Result types are used.
Also removes a compatibility macro in SimpleChannel.cpp, and an identical
specialization in AddonManagerStartup, which are no longer necessary after
this change.
MozReview-Commit-ID: 94iNrPDJEnN
--HG--
extra : rebase_source : 24ad4a54cbd170eb04ded21794530e56b1dfbd82
When used as an error value, nsresult should never be NS_OK, which means that
we should be able to safely pack simple nsresult Result values into a single
word.
MozReview-Commit-ID: GJvnyTPjynk
--HG--
extra : rebase_source : ab5a64b545dfbfe9bbef167f8b63ecbf00b16e07
We now have jemalloc_ptr_info() and moz_malloc_enclosing_size_of(), which can
be used to measure heap blocks via interior pointers. This patch does the
following.
- Adds MOZ_DEFINE_MALLOC_ENCLOSING_SIZE_OF, for defining
measure-via-interior-pointer functions.
- Uses these functions to replace some horrid pointer arithmetic in functions
measuring Rust types.
--HG--
extra : rebase_source : 81940a72a4b6355b1e99cede4db40db68d4afad3
We already periodically calculate the ghost window amount after cycle
collection, this just uses a cached value of that for the distinguished amount.
This avoids the overhead of a recalculating the value when reporting telemetry.
--HG--
extra : histedit_source : 4ba3ee62fd2871a87970faca8a70b2284e83981d%2C9246ac39cea50c3bb0d1693d8e831c3b3ad33ad9
Replace it with NS_INTERFACE_MAP_BEGIN_CYCLE_COLLECTION, because it
has been the same for a while.
MozReview-Commit-ID: 5agRGFyUry1
--HG--
extra : rebase_source : 5388c56b2f6905c6ef969150f0c5b77bf247624d
It appears some Android devices like to send spurious RT signals to our process
which we interpret to mean a gc/cc log is being requested. This causes large
files to pile up in the device storage. We can just disable this feature for
Android as it would be pretty hard for a user to actually use (they can just
go to about:memory). Automation can still enable the FIFO queue if we ever
want to start dumping memory reports on device again.
--HG--
extra : rebase_source : 80cb04473677db7f3edc9377b2fca4c72eb63b71
The virtual call to this method shows up in profiles, and therefore
would be nice to avoid if possible. We can do that by making
nsIWeakReference hold the weak reference itself and therefore
implementing a non-virtual QueryReferent method.
nsQueryReferent is defined as an nsCOMPtr_helper, which implies that
calling its operator() method requires a virtual call. While
nsQueryReferent is marked `final`, compiler inlining decisions make it
impossible to de-virtualize the call to operator(). However, we have
many other classes returned by do_* functions that nsCOMPtr handles
directly, requiring no extra virtual calls, and we can give
nsQueryReferent the same treatment.
We should not be declaring forward declarations for nsString classes directly,
instead we should use nsStringFwd.h. This will make changing the underlying
types easier.
--HG--
extra : rebase_source : b2c7554e8632f078167ff2f609392e63a136c299
This function takes up quite a bit of space, and there's no need to for
the log getter to be inlined everywhere. Moving this to out-of-line
code saves ~200K on x86-64 Linux.
The current definition of MOZ_LOG requires calling
LazyLogModule::operator() multiple times, which is unnecessary. We can
avoid that with a bit of clever macro definition and a lengthy
explanation.
As our threattype-listname conversion design, "goog-harmful-proto" is allocated
for this new threat type. This threat type is mainly for mobile.
MozReview-Commit-ID: G9GbgmHHHfp
--HG--
extra : rebase_source : 0681fcd9322b94451a86eafe57bf1ccc4b89db30
extra : intermediate-source : 28b0502d9add81beeae58a2c33f9fd5839d4d544
extra : source : 646f02f15131aa98ad37015b0a641304a3271796
MozStackWalk() is different on Windows to the other platforms. It has two extra
arguments, which can be used to walk the stack of a different thread.
This patch makes those differences clearer. Instead of having a single function
and forbidding those two arguments on non-Windows, it removes those arguments
from MozStackWalk, and splits off MozStackWalkThread() which retains them. This
also allows those arguments to have more appropriate types (HANDLE instead of
uintptr_t; CONTEXT* instead of than void*) and names (aContext instead of
aPlatformData).
The patch also removes unnecessary reinterpret_casts for the aClosure argument
at a couple of MozStackWalk() callsites.
--HG--
extra : rebase_source : 111ab7d6426d7be921facc2264f6db86c501d127
This callback is only used in very limited ways, so just require that
the caller pass in the canonical supports pointer, plus the
participant. This probably won't affect performance much.
MozReview-Commit-ID: CsThzFsKyYx
--HG--
extra : rebase_source : 9595b1d75fc45bc5ee6d932a840e98b5d760cb78
The nsISupports objects added to the purple buffer are already
canonical, so we can avoid some overhead by not QIing them to
nsCycleCollectionISupports. This QIing overhead shows up in profiles.
MozReview-Commit-ID: CQN6wwc7MZm
--HG--
extra : rebase_source : 27e6b70f83b42b5db7af3d1e7d62e36d6f4013a0
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.
--HG--
extra : rebase_source : 2c23285f8b6c3b667560a9d14014efc4633aed51
This is basically a cosmetic change; references are the normal way to do string
outparams.
--HG--
extra : rebase_source : ffc5945f269bdcd3d4116755b56713e87a44b6cd
As well as the straightforward things, this lets us remove ReadSysFile and
WriteSysFile, which in turn lets us remove TestFileUtils.cpp.
--HG--
extra : rebase_source : fc90c05352e654ffc41009d8504a9c54f394fc3f
We already periodically calculate the ghost window amount after cycle
collection, this just uses a cached value of that for the distinguished amount.
This avoids the overhead of a recalculating the value when reporting telemetry.
This allows a bunch of other things to be removed too, including
PluginModuleParent::mSurrogateInstances,
PluginModuleChromeParent::sInstantiated, and NS_PLUGIN_INIT_PENDING.
The patch also removes the AsyncPluginInit crash annotation.
--HG--
extra : rebase_source : cadb1d215fd93051c9032ea0a1fb6f1d2fb80c6d
Implement nsMemoryReporterManager::GetHeapAllocatedAsync() by dispatching
nsMemoryReporterManager::GetHeapAllocated() to a thread pool and dispatching
the result back to the main thread to run the callback with the requested
attribute.
MozReview-Commit-ID: 9jj1UYqu5KD
--HG--
extra : rebase_source : a3620c6313b3253c41adff8d67c77a347240891b
extra : amend_source : 1732a5e4390fadbc597b6a61dc7d51e892f8441e
extra : histedit_source : b892d2046a07fe00368e8f1976394dc1f4c2f496
nsIMemoryReporter::getHeapAllocatedAsync() is added to get attribute
'heapAllocated' asynchronously.
MozReview-Commit-ID: 96KyZpCeTG1
--HG--
extra : rebase_source : 5c27b26788a3a96821ce054911dda8ce2932007d
extra : amend_source : fcb923dac2dfa9d7aa5bf520d38c5bb11f0d60e9
extra : histedit_source : 83ae65c153f7cef6c6529296838a9d9aa367219e
This requires:
- Moving the constructors of ProfilerMarkerPayload and its subclasses into the
.h file so they are visible even when ProfilerMarkerPayload.cpp isn't
compiled.
- Similarly, using a macro to make StreamPayload() a crashing no-op when the
profiler isn't enabled. (It is never called in that case.)
--HG--
extra : rebase_source : 7aad2fdb1bd4e49782024dba6664e8f992771520
It's silly to use prmem.h within Firefox code given that in our configuration
its functions are just wrappers for malloc() et al. (Indeed, in some places we
mix PR_Malloc() with free(), or malloc() with PR_Free().)
This patch removes all uses, except for the places where we need to use
PR_Free() to free something allocated by another NSPR function; in those cases
I've added a comment explaining which function did the allocation.
--HG--
extra : rebase_source : 0f781bca68b5bf3c4c191e09e277dfc8becffa09
This patch makes the following changes to the macros.
- Removes PROFILER_LABEL_FUNC. It's only suitable for use in functions outside
classes, due to PROFILER_FUNCTION_NAME not getting class names, and it was
mostly misused.
- Removes PROFILER_FUNCTION_NAME. It's no longer used, and __func__ is
universally available now anyway.
- Combines the first two string literal arguments of PROFILER_LABEL and
PROFILER_LABEL_DYNAMIC into a single argument. There was no good reason for
them to be separate, and it forced a '::' in the label, which isn't always
appropriate. Also, the meaning of the "name_space" argument was interpreted
in an interesting variety of ways.
- Adds an "AUTO_" prefix to PROFILER_LABEL and PROFILER_LABEL_DYNAMIC, to make
it clearer they construct RAII objects rather than just being function calls.
(I myself have screwed up the scoping because of this in the past.)
- Fills in the 'js::ProfileEntry::Category::' qualifier within the macro, so
the caller doesn't need to. This makes a *lot* more of the uses fit onto a
single line.
The patch also makes the following changes to the macro uses (beyond those
required by the changes described above).
- Fixes a bunch of labels that had gotten out of sync with the name of the
class and/or function that encloses them.
- Removes a useless PROFILER_LABEL use within a trivial scope in
EventStateManager::DispatchMouseOrPointerEvent(). It clearly wasn't serving
any useful purpose. It also serves as extra evidence that the AUTO_ prefix is
a good idea.
- Tweaks DecodePool::SyncRunIf{Preferred,Possible} so that the labelling is
done within them, instead of at their callsites, because that's a more
standard way of doing things.
--HG--
extra : rebase_source : 318d1bc6fc1425a94aacbf489dd46e4f83211de4
PROFILER_MARKER is now just a trivial wrapper for profiler_add_marker(). This
patch removes it.
--HG--
extra : rebase_source : 9858f34763bb343757896a91ab7ad8bd8e56b076
This patch reduces the differences between builds where the profiler is enabled
and those where the profiler is disabled. It does this by removing numerous
MOZ_GECKO_PROFILER checks.
These changes have the following consequences.
- Various functions and classes are now defined in all builds, and so can be
used unconditionally: profiler_add_marker(), profiler_set_js_context(),
profiler_clear_js_context(), profiler_get_pseudo_stack(), AutoProfilerLabel.
(They are effectively no-ops in non-profiler builds, of course.)
- The no-op versions of PROFILER_* are now gone. The remaining versions are
almost no-ops when the profiler isn't built.
--HG--
extra : rebase_source : 8fb5e8757600210c2f77865694d25162f0b7698a
mozilla::EditorBase stores nsISelectionController and nsIDocument with nsWeakPtr. However, nsWeakPtr requires QI to retrieve actual pointer and it makes some damage to the performance. If mozilla::WeakPter were available for them, it'd be great. However, it's not available with nsISelectionController nor nsIDocument because it's possible to implement SupportsWeakPtr only with their subclasses but the subclasses shouldn't be referred by editor. Therefore, this patch creates mozilla::CachedWeakPtr<class T> which stores both raw pointer (as cache) and nsWeakPtr and when its cache is requested, it checks if the object referred by nsWeakPtr is still alive.
Additionally, this patch hides the members from subclasses of EditorBase for reducing the maintenance cost.
MozReview-Commit-ID: FvtM7453Vv8
--HG--
extra : rebase_source : a524a8ea327c3993645fafa81db8aef65090f1e0
Once the |aPayload| argument to profile_add_marker() became a UniquePtr the
default value of nullptr caused compilation difficulties that could only be
fixed by #including ProfilerMarkerPayload.h into lots of additional places
(because the UniquePtr<T> instantiation required the T to be fully defined). To
get around this I just split profile_add_marker() into two functions, one with
1 argument and one with 2 arguments.
The patch also removes the definition of PROFILER_MARKER_PAYLOAD in the case
where MOZ_GECKO_PROFILER isn't defined. A comment explains why.
Gabriele Svelto
Chris Pearce
Chris Peterson
Nicholas Nethercote
Chia-Hung Duan
Csoregi Natalia
Andrew Osmond
Bevis Tseng
Marco Castelluccio
Kris Maglione
Sylvestre Ledru
Philippe Normand
Stephen A Pohl
Sebastian Hengst
Mike Hommey
Olli Pettay
David Keeler
Kyle Machulis
Andrew McCreight
James Cheng
Tom Ritter
Bobby Holley
Wes Kocher
Eugen Sawin
Nathan Froyd
Eric Rahm
Henri Sivonen
Tristan Bourvon
Ehsan Akhgari
Luke Wagner
Masatoshi Kimura
Ryan VanderMeulen
Henry Chang
Yoshi Huang
Bill McCloskey
Andrea Marchesini
Paul Adenot
Jon Coppeard
Cervantes Yu
David Anderson
Phil Ringnalda
Boris Zbarsky
Mats Palmgren
Carsten "Tomcat" Book
Masayuki Nakano