Luke Wagner
7db2d34bde
Merge
2010-06-21 16:30:17 -07:00
Luke Wagner
ebb2d4890b
Be careful when using setDouble that we shouldn't use setNumber instead to fit in an int when possible. Was missing a lot in jsdate.cpp
2010-06-21 16:28:02 -07:00
David Mandelin
4ab1d47307
Bug 573578: intern string before converting to jsid in PropertyOpForwarder
2010-06-21 16:07:31 -07:00
Luke Wagner
6cc55bb8d5
Fix C-only compile errors
2010-06-21 12:49:23 -07:00
Luke Wagner
5ced30e1fc
Change 32-bit jsval rep to use dense type index instead of bitmask
2010-06-21 12:28:56 -07:00
Luke Wagner
dc916dafc5
Tweak trace-tests/tests/basic/delete-named-names.js checkStats
2010-06-18 01:06:48 -07:00
Luke Wagner
cc9eacee25
Merge
2010-06-18 01:02:26 -07:00
Luke Wagner
bf3afea5b8
Fix bugs; browser is up and running
2010-06-18 00:59:10 -07:00
Luke Wagner
b7955fe7b7
Browser builds
2010-06-17 23:51:54 -07:00
David Mandelin
1ec5c13d2a
Remove MAP_32BIT which prevents us from building on many platforms and we aren't going to use any more anyway.
2010-06-17 18:23:47 -07:00
David Mandelin
c046cbd662
Bug 572842: restore deleted test in ~XPCTraceableVariant
2010-06-17 18:06:24 -07:00
Sean Stangl
bdffe4e8ba
Redefine JSVAL type constants for one-comparison object, primitive check.
...
The following types are defined to have the upper mask bits set:
JSVAL_MASK16_NULL is 0x2000;
JSVAL_MASK16_NONFUNOBJ is 0x4000;
JSVAL_MASK16_FUNOBJ is 0x8000.
So each of IS_OBJECT(), IS_OBJECT_OR_NULL(), and IS_PRIMITIVE()
just requires one comparison, saving a bitwise AND.
2010-06-17 16:44:10 -07:00
Sean Stangl
9a415dd80c
Redefine JSVAL_MASK16_INT32 to ((uint16)0x0001).
...
This permits isNumber() to just check <= JSVAL_MASK32_INT32,
since no number can have type JSVAL_MASK32_CLEAR (asserted).
2010-06-17 16:44:06 -07:00
Luke Wagner
40aca2d2fb
Merge
2010-06-16 23:57:45 -07:00
Luke Wagner
11556d9a6f
Fix bugs introduced by jsid change
2010-06-16 22:58:31 -07:00
Luke Wagner
ab09c2952d
Switch jsids back to pointer-sized and fix a bunch of broken assumptions
2010-06-16 18:21:34 -07:00
David Mandelin
cded156f8e
Bug 569651: add test case that formerly asserted, but is now WFM
2010-06-16 18:14:37 -07:00
David Mandelin
cf66245838
Bug 572229: atomize strings passed to js_HasNamedProperty
2010-06-15 16:00:16 -07:00
David Mandelin
d352690926
Bug 572042: get tracing+fat values to build in the browser. As part of this I disabled the quickstub traceable natives. According to Andreas, we don't necessarily need them for good perf now that we can call the standard quickstub from trace. This build works on simple JS, but there are still bugs (e.g., Google Maps asserts).
2010-06-15 15:43:41 -07:00
David Mandelin
f2f3967750
Bug 572029: fully specialize for unboxing most numbers
2010-06-14 17:26:59 -07:00
Luke Wagner
e54fb2ea80
Make UnboxDoubleHelper extern to match friend declaration
2010-06-14 14:46:31 -07:00
David Mandelin
94e6c8a2cc
Bug 571625: bring back box-after-unbox optimization for tracing with fatvals
2010-06-14 14:24:30 -07:00
David Mandelin
e645d90105
Build fixes for opt browser
2010-06-14 13:11:59 -07:00
Luke Wagner
e01cff89da
Change NPAPI not to cast to from NPIdentifier and jsval. Fix silly JSObject::resizeDenseArrayElements bug introduced by earlier commit
2010-06-11 18:09:58 -07:00
Luke Wagner
2a2b97cc3f
Merge
2010-06-11 14:19:15 -07:00
Luke Wagner
db5e1acb26
Avoid unnecessary initialization of js::Vector
2010-06-11 14:18:45 -07:00
David Mandelin
eec97abca4
Make sure dense_grow gets inlined into the tracer functions that call it
2010-06-11 11:29:04 -07:00
mozilla@buildmonkey-right.mv.mozilla.com
1da69179fa
Write simpler code in JSObject::resizeDenseArrayElements so GCC does not do really really bad register allocation
2010-06-11 10:34:47 -07:00
David Mandelin
f242431382
Bug 569766: Make tracing work with fat values
2010-06-10 17:20:26 -07:00
Luke Wagner
e6f9490099
Simplify code now that Value is copyable
2010-06-10 10:09:05 -07:00
Luke Wagner
e9442c8c26
Doubles are no longer gc-things, fix testGCChunkAlloc
2010-06-10 09:40:11 -07:00
Luke Wagner
a83311a7c5
Fix build bustage in FF
2010-06-10 00:05:41 -07:00
Luke Wagner
dde2523f3a
Remove assumption that getter objects are functions (can be any callable), fix jsproxy aliasing bug
2010-06-09 23:20:37 -07:00
Luke Wagner
1bb5128ab0
Merge with TraceMonkey
2010-06-09 21:59:18 -07:00
Luke Wagner
27e99d6856
Remove OS X rules.mk hacks and out-of-date assert in JS_GetObjectID
2010-06-08 17:32:51 -07:00
Luke Wagner
706cc2c0e9
Firefox starts up and browses the web
2010-06-08 15:12:18 -07:00
Luke Wagner
6d5a111999
Don't re-define js_ObjectOps in xpconnect. Remove assumptions that private jsvals are ints
2010-06-08 00:01:45 -07:00
Jason Orendorff
6e7f075424
Merge backout.
2010-06-07 20:39:48 -05:00
Jason Orendorff
03f2a11f1a
Merge backout.
2010-06-07 20:39:05 -05:00
Jason Orendorff
120196fad2
Back out changeset 46fed5bdcf44 for now.
2010-06-07 20:38:51 -05:00
Jason Orendorff
987881e65c
Merge backout.
2010-06-07 20:38:30 -05:00
Jason Orendorff
5874315d91
Back out changeset 5f425aecd7ab again. Still has performance issues; 14% in ThisObject hook.
2010-06-07 20:36:57 -05:00
Andreas Gal
be035f8008
Allow reading reserved slots without locking if we know its safe to do so (570404, r=brendan).
2010-06-07 21:35:58 -04:00
Luke Wagner
0d5dd30ea8
Firefox builds
2010-06-07 18:21:51 -07:00
Luke Wagner
f032df96b4
Bug 565157 - typed JS_AddRoot (r=igor)
2010-06-07 17:05:02 -07:00
Luke Wagner
a8883b8520
Merge
2010-06-07 13:58:13 -07:00
Luke Wagner
9814243213
xpconnect builds
2010-06-07 13:57:40 -07:00
David Anderson
5ccdb8922d
Bug 556569 - Trace shallow recursion more robustly. r=jorendorff.
2010-06-07 12:47:24 -05:00
Igor Bukanov
24dee0ab53
bug 557538 - custom GC chunk allocation. r=gal,jorendorff
2010-06-07 11:17:15 +02:00
Nicholas Nethercote
3744861727
Update nanojit-import-rev stamp.
2010-06-06 20:27:01 -07:00