The main reason to not do this would be performance (avoiding the
addref/release), but there are two main mitigating factors:
1) All calls to UnwrapReflectorToISupports that pass in a Web IDL object
already do the addref (and in fact QI). So this only affects the
XPCWrappedNative case.
2) The vast majority of the callers proceed to QI on the pointer anyway, and a
second addref is cheap; it's the first addref after a CC that can be
expensive on a cycle-collected object.
Going through the changes one by one:
* In GlobalObject::GetAsSupports, we do have a change that slightly slows down
precisely in the XPCWrappedNative global case. That's the message managers
and the backstagepass. And this really only affects calls to Web IDL statics
from those globals.
* In UnwrapArgImpl we're talking about a Web IDL method taking an "external
interface" type, and the UnwrapReflectorToISupports call is immediately
followed by QI anyway.
* In UnwrapXPConnectImpl we're talking about the case when we have a
non-WebIDL-object implementation of a Web IDL interface. Again, this is the
message manager globals, for EventTarget. And we have a QI call immediately
after the UnwrapReflectorToISupports.
* In the generated HasInstance hook for EventTarget we will be slightly slower
when the LHS of the instanceof is an XPCWrappedNative. And not much slower,
because again there's an immediate QI.
* In InstallXBLField we're never going to have an XPCWrappedNative as thisObj;
it's always an Element in practice. So this is no more expensive than before.
* In sandbox's GetPrincipalOrSOP we now have an extra addref. But it was
followed by various QIs anyway.
* In XPCConvert::JSValToXPCException we have an extra addref if someone throws
an XPCWrappedNative, which is fairly unlikely; our actual Exception objects
are on Web IDL bindings. Plus we have an immediate QI.
* In xpc::HasInstance we have an extra addred if the LHS of instanceof is an
XPCWrappedNative. But, again, there's an immediated QI after the
UnwrapReflectorToISupports.
* In xpcJSWeakReference::Init we are likely doing an extra addref, but again
immediately followed by QI.
I think it's worth making this change just to remove the footgun and that the
perf impact, if any, is pretty minimal.
Currently we resolve a property by iterating every prefable and check whether it
is enabled. If it is, we linear search the ids that it manages. This patch
changes that to binary searching to find whether the id being resolved is
present first, and checking whether its prefable is enabled only when we find
it. This improves the performance of property resolution, especially when the
property is not present.
The patch stores all the property ids a NativePropertiesN owns in a single array
of PropertyInfo structs. Each struct contains an id and the information needed
to find the corresponding Prefable for the enabled check, as well as the
information needed to find the correct property descriptor in the Prefable. We
also store an array of indices into the PropertyInfo array, sorted by bits of
the corresponding jsid. Given a jsid, this allows us to binary search for the
index of the corresponding PropertyInfo, if any. The index array requires 2
bytes for each property, which is ~20k across all our bindings. The extra
information stored in each PropertyInfo requires 4 bytes for each property,
which is about 40k across all our bindings in 32-bit builds, or 80k in 64-bit
builds due to alignment requirements on PropertyInfo. However we save a bit of
memory from changing NativePropertiesN's trios to duos.
The array of unsorted ids is kept because XrayOwnPropertyKeys() includes only
properties that are enabled. Without it, we will need to check every single
property to know whether its prefable is enabled or not, which is inefficient.
With this patch, initializing property ids takes longer because of the sorting.
I measured also insertion sort because I thought the ids should be nearly sorted
as they are generated sequentially at run time, but that's not the case and
NS_QuickSort() runs faster.
MozReview-Commit-ID: Lc4Z1ui3t0o
--HG--
extra : rebase_source : 314efe467a14428c57f90af2ecc0ec5c47a31993
The patch makes the following proxy changes:
* The number of slots in ProxyValueArray is now dynamic and depends on the number of reserved slots we get from the Class.
* "Extra slots" was renamed to "Reserved slots" to make this clearer.
* All proxy Classes now have 2 reserved slots, but it should be easy to change that for proxy Classes that need more than 2 slots.
* Proxies now store a pointer to these slots and this means GetReservedSlot and SetReservedSlot can be used on proxies as well. We no longer need GetReservedOrProxyPrivateSlot and SetReservedOrProxyPrivateSlot.
And some changes to make DOM Proxies work with this:
* We now store the C++ object in the first reserved slot (DOM_OBJECT_SLOT) instead of in the proxy's private slot. This is pretty nice because it matches what we do for non-proxy DOM objects.
* We now store the expando in the proxy's private slot so I removed GetDOMProxyExpandoSlot and changed the IC code to get the expando from the private slot instead.
Specifically, three changes:
1) valueOf should be non-enumerable.
2) valueOf should be === to Object.prototype.valueOf.
3) There should be no toJSON.
The tests come directly from https://github.com/w3c/web-platform-tests/pull/4623
so not much need to review them.
When property getter is specified the desc's value must the set to undefined --
the desc can be re-used by a caller, so it must be done to preserve this
PropertyDescriptor invariant. We also do this in NativeGetOwnPropertyDescriptor.
MozReview-Commit-ID: 39MCwplbzL3
--HG--
extra : rebase_source : c39e3c6808bed3bcb8a1aa9a8194a89e58338a55
The patch is generated from following command:
rgrep -l unused.h|xargs sed -i -e s,mozilla/unused.h,mozilla/Unused.h,
MozReview-Commit-ID: AtLcWApZfES
--HG--
rename : mfbt/unused.h => mfbt/Unused.h
js::Class op are often all null. And when they're not all null, they're often
duplicated among classes. By pulling them out into their own struct, and using a
(possibly null) pointer in js::Class, we can save 114 KiB per process on
64-bit, and half that on 32-bit.
* * *
imported patch separate-ClassOps-2
--HG--
extra : rebase_source : bd751bf247e9491c1966a123dbeffa573657dfb1
This patch makes NativeProperties variable-length and reduces static data by
110,336 bytes on 64-bit, and half that on 32-bit.
MozReview-Commit-ID: 2etZ5AnEhgO
--HG--
extra : rebase_source : 6a167b64df7da3c6940114782fe08337f04a694d
js::ObjectOps is often all null. When it's not all null, it's often duplicated
many times among classes. By pulling it out into its own struct, and using a
(possibly null) pointer in js::Class, we can save 208 KiB per process on
64-bit, and half that on 32-bit.
--HG--
extra : rebase_source : 5be8fe45f652392571b8a6d7b63777cbafba6ae4
nsWrapperCache expects the object it stores to have an ObjectMoved op that will
notify the wrapper cache when the object is moved. SpiderMonkey promises don't
have a way to do this.
The XPCConvert changes are needed to allow code that passes around Promise
objects as nsISupports to continue working instead of ending up with
double-wrapped nsISupports (XPCWrappedNative for an nsISupports XPCWrappedJS)
around the SpiderMonkey Promise.
This is not quite as strong as being able to assert that all codepaths that might
lead to failure call one of the above methods, but being able to assert that
would involve a lot of extra noise at callsites. Or at least changing the
signature of StealNSResult to use an outparam and return a boolean indicating
whether the ErrorResult was failure or not, or something, so StealNSResult
can be usefully called on successful ErrorResults too.
The bulk of this commit was generated with a script, executed at the top
level of a typical source code checkout. The only non-machine-generated
part was modifying MFBT's moz.build to reflect the new naming.
CLOSED TREE makes big refactorings like this a piece of cake.
# The main substitution.
find . -name '*.cpp' -o -name '*.cc' -o -name '*.h' -o -name '*.mm' -o -name '*.idl'| \
xargs perl -p -i -e '
s/nsRefPtr\.h/RefPtr\.h/g; # handle includes
s/nsRefPtr ?</RefPtr</g; # handle declarations and variables
'
# Handle a special friend declaration in gfx/layers/AtomicRefCountedWithFinalize.h.
perl -p -i -e 's/::nsRefPtr;/::RefPtr;/' gfx/layers/AtomicRefCountedWithFinalize.h
# Handle nsRefPtr.h itself, a couple places that define constructors
# from nsRefPtr, and code generators specially. We do this here, rather
# than indiscriminantly s/nsRefPtr/RefPtr/, because that would rename
# things like nsRefPtrHashtable.
perl -p -i -e 's/nsRefPtr/RefPtr/g' \
mfbt/nsRefPtr.h \
xpcom/glue/nsCOMPtr.h \
xpcom/base/OwningNonNull.h \
ipc/ipdl/ipdl/lower.py \
ipc/ipdl/ipdl/builtin.py \
dom/bindings/Codegen.py \
python/lldbutils/lldbutils/utils.py
# In our indiscriminate substitution above, we renamed
# nsRefPtrGetterAddRefs, the class behind getter_AddRefs. Fix that up.
find . -name '*.cpp' -o -name '*.h' -o -name '*.idl' | \
xargs perl -p -i -e 's/nsRefPtrGetterAddRefs/RefPtrGetterAddRefs/g'
if [ -d .git ]; then
git mv mfbt/nsRefPtr.h mfbt/RefPtr.h
else
hg mv mfbt/nsRefPtr.h mfbt/RefPtr.h
fi
--HG--
rename : mfbt/nsRefPtr.h => mfbt/RefPtr.h
The new check in ErrorResult::ReportErrorWithMessage() shouldn't be
needed and is just to protect against the possibility of another way
to construct messages being added.
The bulk of this commit was generated by running:
run-clang-tidy.py \
-checks='-*,llvm-namespace-comment' \
-header-filter=^/.../mozilla-central/.* \
-fix