Summary: Really sorry for the size of the patch. It's mostly automatic
s/nsIDocument/Document/ but I had to fix up in a bunch of places manually to
add the right namespacing and such.
Overall it's not a very interesting patch I think.
nsDocument.cpp turns into Document.cpp, nsIDocument.h into Document.h and
nsIDocumentInlines.h into DocumentInlines.h.
I also changed a bunch of nsCOMPtr usage to RefPtr, but not all of it.
While fixing up some of the bits I also removed some unneeded OwnerDoc() null
checks and such, but I didn't do anything riskier than that.
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
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.
GCHashMap will shortly lose its initialized() function, so we need another way
to indicate whether this table has been created. This patch changes it to use a
pointer instead.
--HG--
extra : rebase_source : 05ce9f2938d1fc8373cbac77c0b3bd62633f93d8
nsJSObjWrapper::mJSObj can be a cross-compartment wrapper, so we can't use it with JSAutoRealm. Store a (same-compartment) global and use that instead.
The object could be a CCW here and we want to make it impossible to get a CCW's global. The first call here is equivalent to checking JS_IsGlobalObject and for the second one JS::CurrentGlobalOrNull(cx) preserves behavior because we wrapped the object into the current compartment.
The sNPObjWrappers table can contain entries for dead JSObjects that have not yet been finalised. We need to take care not to trigger mJSObj's read barrier for such entries since that will attempt to expose the object to JS and cause this assertion.
The patch does this by calling unbarrieredGetPtr() which avoids the barrier.
Note converting a TenuredHeap to bool and testing equality against a pointer don't trigger the barrier.
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
Jon Coppeard
Sylvestre Ledru
arthur.iakab
Steve Fink
Yoshi Cheng-Hao Huang
Shivam Singhal
Boris Zbarsky
Gurzau Raul
Ciure Andrei
Emilio Cobos Álvarez
Jeff Walden
Nicholas Nethercote
Henri Sivonen
Jan de Mooij
Tom Schuster
Jason Orendorff
Brian Hackett
Tom Tromey