Bug 123809 – Build failure when disabling JIT, YARR_JIT, and ASSEMBLER

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Emilio Pozuelo Monfort]

On platforms that don't support MacroAssembler (e.g. powerpc), we disable it (and JIT/YARR_JIT). This worked fine until 2.3.1 which fails to build. I suppose this is a missing #if ENABLE(ASSEMBLER) or similar, but I haven't investigated deeper.

The build failure is:

g++-4.8 -DHAVE_CONFIG_H -I. -I..  -Wall -W -Wcast-align -Wchar-subscripts -Wreturn-type -Wformat -Wformat-security -Wno-format-y2k -Wundef -Wmissing-format-attribute -Wpointer-arith -Wwrite-strings -Wno-unused-parameter -Wno-parentheses -fno-exceptions -DBUILDING_CAIRO__ -DBUILDING_GTK__   -I../Source -I../Source/JavaScriptCore -I../Source/JavaScriptCore/API -I../Source/JavaScriptCore/ForwardingHeaders -I../Source/JavaScriptCore/assembler -I../Source/JavaScriptCore/bytecode -I../Source/JavaScriptCore/bytecompiler -I../Source/JavaScriptCore/debugger -I../Source/JavaScriptCore/dfg -I../Source/JavaScriptCore/disassembler -I../Source/JavaScriptCore/ftl -I../Source/JavaScriptCore/heap -I../Source/JavaScriptCore/interpreter -I../Source/JavaScriptCore/jit -I../Source/JavaScriptCore/llint -I../Source/JavaScriptCore/parser -I../Source/JavaScriptCore/profiler -I../Source/JavaScriptCore/runtime -I../Source/JavaScriptCore/tools -I../Source/JavaScriptCore/yarr -I./DerivedSources/JavaScriptCore -I../Source/WTF  -D_FORTIFY_SOURCE=2 -DENABLE_JIT=0 -DENABLE_YARR_JIT=0 -DENABLE_ASSEMBLER=0  -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security -Wall -Wl,--as-needed -gstabs -pthread -std=c++11 -Wno-c++11-compat -O2 -D_FORTIFY_SOURCE=2 -MT Source/JavaScriptCore/llint/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.o -MD -MP -MF Source/JavaScriptCore/llint/.deps/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.Tpo -c -o Source/JavaScriptCore/llint/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.o `test -f 'Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp' || echo '../'`Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp
In file included from ../Source/JavaScriptCore/bytecode/ValueRecovery.h:30:0,
                 from ../Source/JavaScriptCore/bytecode/CodeOrigin.h:32,
                 from ../Source/JavaScriptCore/bytecode/CodeBlock.h:41,
                 from ../Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp:29:
../Source/JavaScriptCore/jit/GPRInfo.h:201:43: error: expected ')' before 'address'
     JSValueSource(MacroAssembler::Address address)
                                           ^
../Source/JavaScriptCore/jit/GPRInfo.h:269:5: error: 'Address' in 'class MacroAssembler' does not name a type
     MacroAssembler::Address asAddress(unsigned additionalOffset = 0) const { return MacroAssembler::Address(base(), offset() + additionalOffset); }
     ^
make[1]: *** [Source/JavaScriptCore/llint/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.o] Error 1


Full build log: https://buildd.debian.org/status/fetch.php?pkg=webkitgtk&arch=powerpc&ver=2.3.1-1&stamp=1383673445

[Csaba Osztrogonác]

Are you sure if it is an issue on ToT WebKit?
Here is a similar bug report: https://bugs.webkit.org/show_bug.cgi?id=123317

[Emilio Pozuelo Monfort]

I haven't tested with tot, I'll do so beginning of next week and report back.

[Iain Lane]

I just tested trunk & got a similar failure:

g++-4.8 -DHAVE_CONFIG_H -I.  -Wall -W -Wcast-align -Wchar-subscripts -Wreturn-type -Wformat -Wformat-security -Wno-format-y2k -Wundef -Wmissing-format-attribute -Wpointer-arith -Wwrite-strings -Wno-unused-parameter -Wno-parentheses -fno-exceptions -DBUILDING_CAIRO__ -DBUILDING_GTK__   -I./Source -I./Source/JavaScriptCore -I./Source/JavaScriptCore/API -I./Source/JavaScriptCore/ForwardingHeaders -I./Source/JavaScriptCore/assembler -I./Source/JavaScriptCore/bytecode -I./Source/JavaScriptCore/bytecompiler -I./Source/JavaScriptCore/debugger -I./Source/JavaScriptCore/dfg -I./Source/JavaScriptCore/disassembler -I./Source/JavaScriptCore/ftl -I./Source/JavaScriptCore/heap -I./Source/JavaScriptCore/interpreter -I./Source/JavaScriptCore/jit -I./Source/JavaScriptCore/llint -I./Source/JavaScriptCore/parser -I./Source/JavaScriptCore/profiler -I./Source/JavaScriptCore/runtime -I./Source/JavaScriptCore/tools -I./Source/JavaScriptCore/yarr -I./DerivedSources/JavaScriptCore -I./Source/WTF  -D_FORTIFY_SOURCE=2 -DENABLE_JIT=0 -DENABLE_YARR_JIT=0 -DENABLE_ASSEMBLER=0  -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security -Wall -Wl,--as-needed -gstabs -pthread -std=c++11 -Wno-c++11-compat -O2 -D_FORTIFY_SOURCE=2 -MT Source/JavaScriptCore/llint/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.o -MD -MP -MF Source/JavaScriptCore/llint/.deps/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.Tpo -c -o Source/JavaScriptCore/llint/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.o `test -f 'Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp' || echo './'`Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp
In file included from ./Source/JavaScriptCore/bytecode/ValueRecovery.h:30:0,
                 from ./Source/JavaScriptCore/bytecode/CodeOrigin.h:32,
                 from ./Source/JavaScriptCore/bytecode/CodeBlock.h:42,
                 from Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp:29:
./Source/JavaScriptCore/jit/GPRInfo.h:201:43: error: expected ‘)’ before ‘address’
     JSValueSource(MacroAssembler::Address address)
                                           ^
./Source/JavaScriptCore/jit/GPRInfo.h:269:5: error: ‘Address’ in ‘class MacroAssembler’ does not name a type
     MacroAssembler::Address asAddress(unsigned additionalOffset = 0) const { return MacroAssembler::Address(base(), offset() + additionalOffset); }
     ^
In file included from ./Source/WTF/wtf/PossiblyNull.h:29:0,
                 from ./Source/WTF/wtf/FastMalloc.h:27,
                 from ./Source/JavaScriptCore/config.h:60,
                 from Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp:26:
./Source/JavaScriptCore/jit/GPRInfo.h:772:16: error: ‘GPRInfo’ has not been declared
 COMPILE_ASSERT(GPRInfo::regT0 == GPRInfo::returnValueGPR, regT0_must_equal_returnValueGPR);
                ^
./Source/WTF/wtf/Assertions.h:325:50: note: in definition of macro ‘COMPILE_ASSERT’
 #define COMPILE_ASSERT(exp, name) static_assert((exp), #name)
                                                  ^
./Source/JavaScriptCore/jit/GPRInfo.h:772:34: error: ‘GPRInfo’ has not been declared
 COMPILE_ASSERT(GPRInfo::regT0 == GPRInfo::returnValueGPR, regT0_must_equal_returnValueGPR);
                                  ^
./Source/WTF/wtf/Assertions.h:325:50: note: in definition of macro ‘COMPILE_ASSERT’
 #define COMPILE_ASSERT(exp, name) static_assert((exp), #name)
                                                  ^
./Source/JavaScriptCore/jit/GPRInfo.h:774:16: error: ‘GPRInfo’ has not been declared
 COMPILE_ASSERT(GPRInfo::regT1 == GPRInfo::returnValueGPR2, regT1_must_equal_returnValueGPR2);
                ^
./Source/WTF/wtf/Assertions.h:325:50: note: in definition of macro ‘COMPILE_ASSERT’
 #define COMPILE_ASSERT(exp, name) static_assert((exp), #name)
                                                  ^
./Source/JavaScriptCore/jit/GPRInfo.h:774:34: error: ‘GPRInfo’ has not been declared
 COMPILE_ASSERT(GPRInfo::regT1 == GPRInfo::returnValueGPR2, regT1_must_equal_returnValueGPR2);
                                  ^
./Source/WTF/wtf/Assertions.h:325:50: note: in definition of macro ‘COMPILE_ASSERT’
 #define COMPILE_ASSERT(exp, name) static_assert((exp), #name)
                                                  ^
make: *** [Source/JavaScriptCore/llint/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.o] Error 1

[Emilio Pozuelo Monfort]

Still an issue, see https://buildd.debian.org/status/logs.php?pkg=webkitgtk&ver=2.3.2-1&suite=experimental

[Mark Lam]

Created attachment 217751 [details]
the patch.

[EFL EWS Bot]

Comment on attachment 217751 [details]
the patch.

Attachment 217751 [details] did not pass efl-ews (efl):
Output: http://webkit-queues.appspot.com/results/34908066

[Geoffrey Garen]

Comment on attachment 217751 [details]
the patch.

r=me

[Build Bot]

Comment on attachment 217751 [details]
the patch.

Attachment 217751 [details] did not pass win-ews (win):
Output: http://webkit-queues.appspot.com/results/34948065

[EFL EWS Bot]

Comment on attachment 217751 [details]
the patch.

Attachment 217751 [details] did not pass efl-wk2-ews (efl-wk2):
Output: http://webkit-queues.appspot.com/results/34998074

[Filip Pizlo]

I think that this is the wrong approach. 

tryToDisassemble should be defined even if !ENABLE(JIT). This would allow you to get rid of some of these #if's. 

In the future when these bugs arise I think it's better to add stubs for !ENABLE(JIT) than to start splattering #if's all over the codebase.

[kov's GTK+ EWS bot]

Comment on attachment 217751 [details]
the patch.

Attachment 217751 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/34918091

[Iain Lane]

Gets further with this, but eventually hits

libtool: link: g++-4.8 -pthread -std=c++11 -Wno-c++11-compat -O2 -D_FORTIFY_SOURCE=2 -Wl,-z -Wl,relro -Wl,--no-keep-memory -Wl,--no-demangle -o Programs/LLIntOffsetsExtractor Source/JavaScriptCore/llint/Programs_LLIntOffsetsExtractor-LLIntOffsetsExtractor.o -Wl,--export-dynamic -pthread  ./.libs/libWTF.a -lz -licui18n -licuuc -licudata -lgmodule-2.0 -lgthread-2.0 -lgio-2.0 -lgobject-2.0 -lglib-2.0 -lm -lpthread -lstdc++ -pthread
/usr/bin/ruby ./Source/JavaScriptCore/offlineasm/asm.rb ./Source/JavaScriptCore/llint/LowLevelInterpreter.asm Programs/LLIntOffsetsExtractor DerivedSources/JavaScriptCore/LLIntAssembly.h
offlineasm: Parsing ./Source/JavaScriptCore/llint/LowLevelInterpreter.asm and Programs/LLIntOffsetsExtractor and creating assembly file DerivedSources/JavaScriptCore/LLIntAssembly.h.
offlineasm: Including file ./Source/JavaScriptCore/llint/LowLevelInterpreter64.asm
offlineasm: Including file ./Source/JavaScriptCore/llint/LowLevelInterpreter32_64.asm
/home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:393:in `validate': Unresolved extraStackSpace at ./Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:91 (RuntimeError)
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:399:in `block in validateChildren'
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:397:in `each'
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:397:in `validateChildren'
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:456:in `validate'
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:399:in `block in validateChildren'
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:397:in `each'
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:397:in `validateChildren'
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/transform.rb:406:in `validate'
	from ./Source/JavaScriptCore/offlineasm/asm.rb:268:in `block (3 levels) in <main>'
	from /home/laney/WebKit/Source/JavaScriptCore/offlineasm/settings.rb:89:in `forSettings'
	from ./Source/JavaScriptCore/offlineasm/asm.rb:265:in `block (2 levels) in <main>'
	from ./Source/JavaScriptCore/offlineasm/asm.rb:261:in `each'
	from ./Source/JavaScriptCore/offlineasm/asm.rb:261:in `block in <main>'
	from ./Source/JavaScriptCore/offlineasm/asm.rb:252:in `open'
	from ./Source/JavaScriptCore/offlineasm/asm.rb:252:in `<main>'
make: *** [DerivedSources/JavaScriptCore/LLIntAssembly.h] Error 1

That's when configuring like so:

CPPFLAGS="-D_FORTIFY_SOURCE=2 -DENABLE_JIT=0 -DENABLE_YARR_JIT=0 -DENABLE_ASSEMBLER=0" LDFLAGS="-Wl,-z,relro -Wl,--no-keep-memory" CC="gcc-4.8" CXX="g++-4.8" ./configure --prefix=/usr --with-gtk=2.0 --disable-silent-rules --host=powerpc-linux-gnu --build=powerpc-linux-gnu --enable-gtk-doc --enable-introspection --enable-geolocation --disable-webkit2

[Mark Lam]

Created attachment 218206 [details]
patch 2: Need to fix the build when disabling the DISASSEMBLER as well.

[Mark Lam]

Comment on attachment 218206 [details]
patch 2: Need to fix the build when disabling the DISASSEMBLER as well.

The EWS bots look good.  Time for a review.

[Geoffrey Garen]

Comment on attachment 218206 [details]
patch 2: Need to fix the build when disabling the DISASSEMBLER as well.

r=me

[Mark Lam]

Landed in r159987: <http://trac.webkit.org/r159987>.

[Tomas Popela]

Should the proposed patch fix the issue mentioned in comment #12? If so I'm still getting it on ppc32 (ppc64 is working fine)..

[Mark Lam]

(In reply to comment #17)
> Should the proposed patch fix the issue mentioned in comment #12? If so I'm still getting it on ppc32 (ppc64 is working fine)..

It builds fine for me.  Can you confirm that you are building with ENABLE_LLINT_C_LOOP set to 1?

[Iain Lane]

(In reply to comment #18)
> (In reply to comment #17)
> > Should the proposed patch fix the issue mentioned in comment #12? If so I'm still getting it on ppc32 (ppc64 is working fine)..
> 
> It builds fine for me.  Can you confirm that you are building with ENABLE_LLINT_C_LOOP set to 1?

I think that's forced on by disabling ENABLE_JIT, isn't it?

[Mark Lam]

(In reply to comment #19)
> (In reply to comment #18)
> > (In reply to comment #17)
> > > Should the proposed patch fix the issue mentioned in comment #12? If so I'm still getting it on ppc32 (ppc64 is working fine)..
> > 
> > It builds fine for me.  Can you confirm that you are building with ENABLE_LLINT_C_LOOP set to 1?
> 
> I think that's forced on by disabling ENABLE_JIT, isn't it?

Yes, but I didn't know if Tomas' build environment has customizations that changes that or not, and I didn't want to assume.  So, I asked for the verification.

[Tomas Popela]

Actually I was compiling the WebKitGTK+ 2.3.2 (snapshot of WebKit trunk ~ 2 weeks ago) and I applied patches from Mark but I intentionally missed this one https://bugs.webkit.org/show_bug.cgi?id=125186 (as I'm not compiling on Win) but that patch was what I missed. So sorry about that, it's compiling successfully now.

[Iain Lane]

(In reply to comment #21)
> Actually I was compiling the WebKitGTK+ 2.3.2 (snapshot of WebKit trunk ~ 2 weeks ago) and I applied patches from Mark but I intentionally missed this one https://bugs.webkit.org/show_bug.cgi?id=125186 (as I'm not compiling on Win) but that patch was what I missed. So sorry about that, it's compiling successfully now.

Ah, thanks; trying that myself now (last tried trunk when I posted the failure, which was before this change).