Since the ABI on this platform just uses the ARM procedure call
standard, which is the same standard Unix uses, we can just
copy-and-paste everything, with some adjustments to the assembly code to
compile properly.
In many places, nsTArray::IndexOf is followed by accessing that element
(hopefully with `Elements() + index`, which skips unnecessary bounds checks.)
But this pattern introduces operations that could be avoided:
- IndexOf converts the address of the found element into an index,
- The caller must test for a special `NoIndex` value,
- On success, accesses convert that index back into the original address.
This patch introduces new 'ApplyIf...' functions that combine all these
operations in a more efficient ensemble: If the sought element is found, it is
passed by reference to a given callable object (usually a lambda); if not
found, another callable is invoked.
Depending on what they need, the first callable may take one of these parameter
lists: (), (size_t), (maybe-const elem_type&), (size_t, maybe-const elem_type&).
On top of removing the pointer->index->pointer operations in most cases,
invoking callables directly from ApplyIf is safer, as the array is guaranteed to
be untouched at this time.
Also, being templates taking function objects, in most cases the compiler should
be able to inline and optimize the search and its callables' code.
This patch gives example uses in nsTArray::Contains, and in
FrameProperties::GetInternal, SetInternal.
And FrameProperties::Has now calls Contains, which is more efficient because
the former code would compute an index (or NoIndex), and then convert that index
to a bool; whereas the new code directly produces a bool from within the search
algorithm.
Differential Revision: https://phabricator.services.mozilla.com/D2758
--HG--
extra : moz-landing-system : lando
This restores the old allocation semantics for "append" operations between
Latin1 and UTF-16 while keeping the buffer re-use optimization for the
"assign" cases.
MozReview-Commit-ID: 8JCw3AaCNLN
Differential Revision: https://phabricator.services.mozilla.com/D3582
--HG--
extra : moz-landing-system : lando
Shrinking the buffer is purely a memory footprint optimization and can be
omitted as far as the string semantics visible to the caller are concerned.
Since shrinking is optional, it doesn't make sense to propagate error when
it fails due to OOM.
MozReview-Commit-ID: BuyBLCBmYzZ
Differential Revision: https://phabricator.services.mozilla.com/D3866
--HG--
extra : moz-landing-system : lando
As initially implemented, nsITLSServerSocket by default enabled the use of the
TLS session cache provided by NSS. However, no consumers of nsITLSServerSocket
actually used it. Because it was an option, though, PSM had to jump through some
hoops to a) make it work in the first place and b) not have NSS panic on
shutdown. Furthermore, it meant increased memory usage for every user of Firefox
(and again, nothing actually used the feature, so this was for naught).
In bug 1479918, we discovered that if PSM shut down before Necko, NSS could
attempt to acquire a lock on the session cache that had been deleted, causing a
shutdown hang. We probably should make it less easy to make this mistake in NSS,
but in the meantime bug 1479918 needs uplifting and this workaround is the
safest, most straight-forward way to achieve this.
Differential Revision: https://phabricator.services.mozilla.com/D3919
--HG--
extra : moz-landing-system : lando
The 'x' prefix makes it clearer that these are infallible.
A couple of nsJSID methods are now also infallible.
--HG--
extra : rebase_source : fcce44a00212d6d341afbf3827b31bd4f7355ad5
There are surprisingly many of them.
(Plus a couple of unnecessary checks after `new` calls that were nearby.)
--HG--
extra : rebase_source : 47b6d5d7c5c99b1b50b396daf7a3b67abfd74fc1
This makes it much easier to update existing consumers of
XPCOMUtils.enumerateCategoryEntries to use the category manager directly.
It also, unfortunately, requires updating existing category manager consumers
to use the Services getter in order to avoid ESLint errors.
Differential Revision: https://phabricator.services.mozilla.com/D4278
--HG--
extra : rebase_source : fb9fd9b21db80af472ff6250a2e9a35e8d538147
Nearly all of the consumers of category enumerators require the entry value,
either along with or instead of the name. Including both by default simplifies
things considerably for most consumers, and allows us to remove the XPCOMUtils
wrapper that JS callers typically use to enumerate category entries.
Differential Revision: https://phabricator.services.mozilla.com/D4277
--HG--
extra : rebase_source : 1efe0434e150f08bb9f4d8a4c6f6e993efebf8d8
This allows JS callers to automatically get the correct types during
interation, without having to explicitly specify them.
Differential Revision: https://phabricator.services.mozilla.com/D3728
--HG--
extra : rebase_source : b708f382d8ea571d199c669bfed5b5a7ca9ffac4
extra : histedit_source : 7df6feb82088c8a5ca45dc28fe4d2b852c177fee
This patch adds simple stubs to convert between the nsISimpleEnumerator
iteration protocol and the JS iteration protocol.
Each iterable object is required to have an @@iterator method which returns an
object implementing the iterator protocol. The later objects, by convention,
also have such a method which returns the object itself.
This patch adds both a @@iterator() and entries() methods to
nsISimpleEnumerator. The former returns an iterator which returns plain
nsISupports objects. The latter accepts an IID and queries each element to
that IID before returning it. If any element fails to query, the error is
propagated to the caller.
Differential Revision: https://phabricator.services.mozilla.com/D3727
--HG--
extra : rebase_source : 340eb43a1c5e6d7ae69fa8ee486d66d31d079b14
extra : histedit_source : f3efc6c265851a563968ee410e4626e0540f55c0
The nsISimpleEnumerator contract specifies that GetNext() returns
NS_ERROR_FAILURE when iteration is complete. Several implementations, however,
either return NS_OK and a null result, or return some other error code, when
iteration is complete.
Since my initial implementation of the JS iteration stubs rely on the
contract-defined behavior of GetNext(), these need to be fixed before it can
land.
Differential Revision: https://phabricator.services.mozilla.com/D3726
--HG--
extra : rebase_source : aab0395df52e18ccff5b0a2489a983013bf484b1
extra : histedit_source : a5644f0a88799b4463af9dd01dfec33b373b1f58
In order to allow JS callers to use nsISimpleEnumerator instances with the JS
iteration protocol, we'll need to additional methods to every instance. Since
we currently have a large number of unrelated implementations, it would be
best if they could share the same implementation for the JS portion of the
protocol.
This patch adds a stub nsSimpleEnumerator base class, and updates all existing
implementations to inherit from it. A follow-up will add a new base interface
to this class, and implement the additional functionality required for JS
iteration.
Differential Revision: https://phabricator.services.mozilla.com/D3725
--HG--
extra : rebase_source : ad66d7b266856d5a750c772e4710679fab9434b1
extra : histedit_source : a83ebffbf2f0b191ba7de9007f73def6b9a955b8
This patch allows us to define methods or getters/setters for any of the
current set of well-known symbols. Those are defined by adding the [symbol]
attribute to a method:
[symbol]
Iterator iterator();
which causes the method to define a property with the well-known symbol which
matches its method name (Symbol.iterator, in this case).
Due to the implementation details of the XPIDL parser, this currently does not
support defining a non-symbol function with the same name as a symbol
function:
[symbol]
Iterator iterator();
[binaryname(OtherIterator)]
Thing iterator(in nsIDRef aIID);
throws for a duplicate method name, even though there is no actual conflict.
Differential Revision: https://phabricator.services.mozilla.com/D3724
--HG--
extra : rebase_source : 1385e2da93113306730f7c087fe7385dbe668e91
extra : histedit_source : 3afd9fe38e7cbddc5576c2bd1673496dd623e489
This assertion was always meant to be a best-effort thing to catch obvious
errors, but the cases where the assumptions it makes fail have been growing. I
could remove it entirely, but I'd be happier keeping at least some basic
sanity checks.
This compromise continues allowing any address below the first argument
pointer, and extends the assertion to also allow anything more than 2KiB above
it. We could probably get away with stretching that to at least 4KiB, but 2
seems safer, and likely enough to catch the obvious cases.
Differential Revision: https://phabricator.services.mozilla.com/D3542
--HG--
extra : source : a5f51c76930c49160bf5e909301d8e7f1a83e379
extra : amend_source : e3decc44bfb4bed6696a394980c378dc033e1021
This assertion was always meant to be a best-effort thing to catch obvious
errors, but the cases where the assumptions it makes fail have been growing. I
could remove it entirely, but I'd be happier keeping at least some basic
sanity checks.
This compromise continues allowing any address below the first argument
pointer, and extends the assertion to also allow anything more than 2KiB above
it. We could probably get away with stretching that to at least 4KiB, but 2
seems safer, and likely enough to catch the obvious cases.
Differential Revision: https://phabricator.services.mozilla.com/D3542
--HG--
extra : rebase_source : 9e17aa3d751f75deff67e40d223fda7aaa4f84f0
extra : amend_source : 86ada4dd01584defeaa5dcb093ad9a73b34f0318
In order to implement profile-per-install we need a mutable INI parser in early
startup. The current one is implemented in JavaScript and thus not available.
This makes the current read-only C++ INI parser mutable and removes the
JavaScript implementation.
It turns out that the two different implementations of nsIINIParserFactory and
nsIINIParser behaved slightly differently but only in ways that the single test
cared about so I've adjusted things a little to make it work.
The existing C++ implementation did not do validity checks on arguments, this
adds that making empty sections and values illegal.
Differential Revision: https://phabricator.services.mozilla.com/D3851
--HG--
rename : xpcom/tests/unit/test_iniProcessor.js => xpcom/tests/unit/test_iniParser.js
extra : source : 524941c8ed0e048ee51be1bd11082b41428ef490
extra : amend_source : 2de6cef5be97448a41733bedda29d6af34aed27a
This moves the code that detects very low memory scenarios and grabs memory
reports from the main thread event-loop to the available memory tracker.
Besides removing the overhead of the check from the event-loop code this
increases the likeliness of the reports being gathered by sampling at a
higher frequency but only when we already detected a low-memory scenario. Last
but not least this add checks for low commit-space detection alongside low
virtual-memory detection.
Differential Revision: https://phabricator.services.mozilla.com/D3669
--HG--
extra : moz-landing-system : lando
Now that :-moz-system-metric is gone, there's no real reason for the atoms to
be separate.
Differential Revision: https://phabricator.services.mozilla.com/D3497
--HG--
extra : moz-landing-system : lando
Correctness improvements:
* UTF errors are handled safely per spec instead of dangerously truncating
strings.
* There are fewer converter implementations.
Performance improvements:
* The old code did exact buffer length math, which meant doing UTF math twice
on each input string (once for length calculation and another time for
conversion). Exact length math is more complicated when handling errors
properly, which the old code didn't do. The new code does UTF math on the
string content only once (when converting) but risks allocating more than
once. There are heuristics in place to lower the probability of
reallocation in cases where the double math avoidance isn't enough of a
saving to absorb an allocation and memcpy.
* Previously, in UTF-16 <-> UTF-8 conversions, an ASCII prefix was optimized
but a single non-ASCII code point pessimized the rest of the string. The
new code tries to get back on the fast ASCII path.
* UTF-16 to Latin1 conversion guarantees less about handling of out-of-range
input to eliminate an operation from the inner loop on x86/x86_64.
* When assigning to a pre-existing string, the new code tries to reuse the
old buffer instead of first releasing the old buffer and then allocating a
new one.
* When reallocating from the new code, the memcpy covers only the data that
is part of the logical length of the old string instead of memcpying the
whole capacity. (For old callers old excess memcpy behavior is preserved
due to bogus callers. See bug 1472113.)
* UTF-8 strings in XPConnect that are in the Latin1 range are passed to
SpiderMonkey as Latin1.
New features:
* Conversion between UTF-8 and Latin1 is added in order to enable faster
future interop between Rust code (or otherwise UTF-8-using code) and text
node and SpiderMonkey code that uses Latin1.
MozReview-Commit-ID: JaJuExfILM9
Because mozilla::HashSet is much faster than PLDHashTable, and mPtrToNodeMap is
hot.
--HG--
extra : rebase_source : bd861a1cd07ae27f4d18ce5c01ca40c5bb73c035
Entry storage allocation now occurs on the first lookupForAdd()/put()/putNew().
This removes the need for init() and initialized(), and matches how
PLDHashTable/nsTHashtable work. It also removes the need for init() functions
in a lot of types that are built on top of mozilla::Hash{Map,Set}.
Pros:
- No need for init() calls and subsequent checks.
- No memory allocated for empty tables, which are not that uncommon.
Cons:
- An extra branch in lookup() and lookupForAdd(), but not in put()/putNew(),
because the existing checkOverloaded() can handle it.
Specifics:
- Construction now can take a length parameter.
- init() is removed. Explicit length-setting, when necessary, now occurs in the
constructors.
- initialized() is removed.
- capacity() now returns zero when the entry storage is absent.
- lookupForAdd() is no longer `const`, because it can instantiate the storage,
which requires modifications.
- lookupForAdd() can now return an invalid AddPtr in two cases:
- old: hashing failure (due to OOM in the hasher)
- new: OOM while instantiating entry storage
The existing failure handling paths for the old case work for the new case.
- clear(), finish(), and clearAndShrink() are replaced by clear(), compact(),
and reserve(). The old compactIfUnderloaded() is also removed.
- Capacity computation code is now in its own functions, bestCapacity() and
hashShift(). setTableSizeLog2() is removed.
- uint32_t is used throughout for capacities, instead of size_t, for
consistency with other similar values.
- changeTableSize() now takes a capacity instead of a deltaLog2, and it can now
handle !mTable.
Measurements:
- Total source code size is reduced by over 900 lines. Also, lots of existing
lines got shorter (i.e. two checks were reduced to one).
- Executable size barely changed, down by 2 KiB on Linux64. The extra branches
are compensated for by the lack of init() calls.
- Speed changed negligibly. The instruction count for Bench_Cpp_MozHash
increased from 2.84 billion to 2.89 billion but any execution time change was
well below noise.
This gcc 4.9 workaround (from bug 1377351) is no longer needed because Firefox currently requires gcc 6.1 or later (as of bug 1444274).
MozReview-Commit-ID: 9R14BDzWEoj
--HG--
extra : rebase_source : a56ec3ee321cdc76e704fe33c2c4a5b85b558889
extra : source : e0c26ec11d499058e51bc2c3d06b2e1840e77f13
This introduces the machinery needed to generate crash annotations from a YAML
file. The relevant C++ functions are updated to take a typed enum. JavaScript
calls are unaffected but they will throw if the string argument does not
correspond to one of the known entries in the C++ enum. The existing whitelists
and blacklists of annotations are also generated from the YAML file and all
duplicate code related to them has been consolidated. Once written out to the
.extra file the annotations are converted in string form and are no different
than the existing ones.
All existing annotations have been included in the list (and some obsolete ones
have been removed) and all call sites have been updated including tests where
appropriate.
--HG--
extra : source : 4f6c43f2830701ec5552e08e3f1b06fe6d045860
ffxbld
Henri Sivonen
Jan de Mooij
Narcis Beleuzu
Brian Hackett
Luke Wagner
Olli Pettay
Paolo Amadini
Nathan Froyd
Margareta Eliza Balazs
Andreea Pavel
Gerald Squelart
Ciure Andrei
David Keeler
Nicholas Nethercote
Kris Maglione
aceman
Landry Breuil
Marco Castelluccio
shindli
James Willcox
Mike Conley
Dave Townsend
Dorel Luca
Gabriele Svelto
Daosheng Mu
Emilio Cobos Álvarez
Tiberius Oros
Ehsan Akhgari
Cameron McCormack
Csoregi Natalia
Diego Pino Garcia
Chris Peterson
Jan Varga
Daniel Varga