Created attachment 639454 [details] [diff] [review] Use payloadOf/PAYLOAD_OFFSET when constructing PIC stubs TenFourFox: Bigger Endian is Better(tm) Bug 750733 introduced handles widely as part of the improved GC, but the current implementation makes assumptions about the word order, causing crashes with things like argument objects on big-endian platforms. See also https://code.google.com/p/tenfourfox/issues/detail?id=165#c39 (note 39 and following). This small patch should work on both little- and big-endian, and passes all tests with the PowerPC backend.
Comment on attachment 639454 [details] [diff] [review] Use payloadOf/PAYLOAD_OFFSET when constructing PIC stubs I think Brian's a better reviewer for this one.
Thanks, David and Brian!
https://hg.mozilla.org/integration/mozilla-inbound/rev/2dd36c1cd662 Bonus points next time if your patches include the needed metadata. It makes life easier for checkin monkeys like me :-) https://developer.mozilla.org/en/Creating_a_patch_that_can_be_checked_in
http://mozillamemes.tumblr.com/post/19498220636/try-server-takes-the-beatings-so-mozilla-inbound Compiling is also a nice feature for a patch to have. Backed out due to red. https://hg.mozilla.org/integration/mozilla-inbound/rev/08fb8d6997bc https://tbpl.mozilla.org/php/getParsedLog.php?id=13334873&tree=Mozilla-Inbound PolyIC.cpp /usr/bin/ccache /tools/gcc-4.5-0moz3/bin/g++ -o PolyIC.o -c -I./../../dist/system_wrappers_js -include /builds/slave/m-in-lnx64/build/js/src/config/gcc_hidden.h -DMOZ_GLUE_IN_PROGRAM -DEXPORT_JS_API -DJS_HAS_CTYPES -DDLL_PREFIX=\"lib\" -DDLL_SUFFIX=\".so\" -DNO_NSPR_10_SUPPORT -Ictypes/libffi/include -I. -I/builds/slave/m-in-lnx64/build/js/src/../../mfbt/double-conversion -I/builds/slave/m-in-lnx64/build/js/src -I. -I./../../dist/include -I/builds/slave/m-in-lnx64/build/obj-firefox/dist/include/nspr -I/builds/slave/m-in-lnx64/build/js/src -I/builds/slave/m-in-lnx64/build/js/src/assembler -I/builds/slave/m-in-lnx64/build/js/src/yarr -fPIC -pedantic -Wall -Wpointer-arith -Woverloaded-virtual -Werror=return-type -Wtype-limits -Wempty-body -Wno-ctor-dtor-privacy -Wno-overlength-strings -Wno-invalid-offsetof -Wno-variadic-macros -Wcast-align -Wno-long-long -fno-rtti -ffunction-sections -fdata-sections -fno-exceptions -pthread -pipe -DNDEBUG -DTRIMMED -g -O3 -freorder-blocks -finline-limit=50 -fno-omit-frame-pointer -DUSE_SYSTEM_MALLOC=1 -DENABLE_ASSEMBLER=1 -DENABLE_JIT=1 -DMOZILLA_CLIENT -include ./js-confdefs.h -MD -MF .deps/PolyIC.o.pp /builds/slave/m-in-lnx64/build/js/src/methodjit/PolyIC.cpp ../../../js/src/methodjit/PunboxAssembler.h: In member function 'void js::mjit::GetPropCompiler::generateGetterStub(js::mjit::Assembler&, js::Shape*, jsid, js::mjit::MacroAssemblerTypedefs::Label, js::Vector<JSC::AbstractMacroAssembler<JSC::X86Assembler>::Jump, 8ul>&)': ../../../js/src/methodjit/PunboxAssembler.h:52:44: error: 'const uint32_t js::mjit::PunboxAssembler::PAYLOAD_OFFSET' is private ../../../js/src/methodjit/PolyIC.cpp:1102:80: error: within this context
FWIW, all the build failures were 64bit.
error: 'const uint32_t js::mjit::PunboxAssembler::PAYLOAD_OFFSET' is private? Should I make that public also? Sorry about the metadata; I'll just generate checkin-ready versions separately.
Actually, I take that back -- it seems easier just to make this 32-bit only. Would that be the right approach, Brian?
(In reply to Cameron Kaiser from comment #7) > Actually, I take that back -- it seems easier just to make this 32-bit only. > Would that be the right approach, Brian? I think that making PAYLOAD_OFFSET public in PunboxAssembler.h should be fine (and seems simple?), it'd be nice to avoid any #ifdef JS_BLAH in the IC methods.
Created attachment 640116 [details] [diff] [review] Use payloadOf/PAYLOAD_OFFSET when constructing PIC stubs (v2) Okay, patched PunboxAssembler.h also. Unfortunately, I do not have access to push to try.
Still missing commit info :-P https://tbpl.mozilla.org/?tree=Try&rev=689f9db78714
Comment on attachment 640116 [details] [diff] [review] Use payloadOf/PAYLOAD_OFFSET when constructing PIC stubs (v2) Pushed to try @ https://tbpl.mozilla.org/?tree=Try&rev=7520fdb6c037
On bhackett's try, I see a single red mochitest for Android and I don't know if that has anything to do with this, plus some oranges under Linux 64bit. OS X64 and Win64 are green. I don't think this is the cause of the fail/oranges.
It builds, good enough for me :) https://hg.mozilla.org/integration/mozilla-inbound/rev/f3145f3e7f44