Bug 60951 – [Qt] bad codegen, pointer diff in JSC::JSCallbackConstructor::JSCallbackConstructor

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Benjamin Reed]

Created attachment 93754 [details]
patch against the qtwebkit 2.0 branch

(I left the version at "Nightly build" since I wasn't sure what versions correspond to the 2.0/2.1/2.2 branches...)

On Mac OS X, when building QtWebkit as a standalone project, compilation fails with an error like os:

---(snip!)---
g++-4.2 -O -c -pipe -I/32sw/lib/system-openssl/include -Wall -Wextra -Wreturn-type -fno-strict-aliasing -Wcast-align -Wchar-subscripts -Wformat-security -Wreturn-type -Wno-unused-parameter -Wno-sign-compare -Wno-switch -Wno-switch-enum -Wundef -Wmissing-noreturn -Winit-self -O2 -D__USE_WS_X11__ -DBUILDING_QT__=1 -DWTF_USE_ACCELERATED_COMPOSITING -DNDEBUG -DBUILDING_QT__ -DBUILDING_JavaScriptCore -DBUILDING_WTF -DQT_NO_DEBUG -DQT_CORE_LIB -DQT_SHARED -I/32sw/lib/qt4-x11/mkspecs/darwin-g++ -I../../../JavaScriptCore -I/32sw/lib/qt4-x11/include/QtCore -I/32sw/lib/qt4-x11/include -I../../../JavaScriptCore -I../../../../qtwebkit-build -I../../../JavaScriptCore/assembler -I../../../JavaScriptCore/bytecode -I../../../JavaScriptCore/bytecompiler -I../../../JavaScriptCore/debugger -I../../../JavaScriptCore/interpreter -I../../../JavaScriptCore/jit -I../../../JavaScriptCore/parser -I../../../JavaScriptCore/pcre -I../../../JavaScriptCore/profiler -I../../../JavaScriptCore/runtime -I../../../JavaScriptCore/wtf -I/private/tmp/qtwebkit-build/JavaScriptCore/wtf/symbian -I../../../JavaScriptCore/wtf/unicode -I../../../JavaScriptCore/yarr -I../../../JavaScriptCore/API -I../../../JavaScriptCore/ForwardingHeaders -Igenerated -I../include/QtWebKit -I. -I../../../JavaScriptCore -I. -o obj/release/jsc.o ../../../JavaScriptCore/jsc.cpp
g++-4.2 -headerpad_max_install_names -L/32sw/lib/system-openssl/lib -prebind -o ./jsc obj/release/jsc.o    -L./release -L/32sw/lib/qt4-x11/lib -ljscore -lQtCore -L/32sw/lib/system-openssl/lib -L/32sw/lib/freetype219/lib -L/32sw/lib/fontconfig2/lib -L/32sw/lib -L/usr/X11R6/lib -L/32sw/lib/qt4-x11/lib -liconv -Wl,-framework,Carbon -Wl,-framework,AppKit -Wl,-framework,ApplicationServices -Wl,-framework,SystemConfiguration -lcrypto -lpng -lresolv -lssl -lz 
ld: bad codegen, pointer diff in JSC::JSCallbackConstructor::JSCallbackConstructor(JSC::JSGlobalObject*, WTF::NonNullPassRefPtr<JSC::Structure>, OpaqueJSClass*, OpaqueJSValue* (*)(OpaqueJSContext const*, OpaqueJSValue*, unsigned long, OpaqueJSValue const* const*, OpaqueJSValue const**))to global weak symbol vtable for JSC::JSCellfor architecture i386
collect2: ld returned 1 exit status
---(snip!)---

This is the error you get in the webkit project in the 2.1 or 2.2 branches, in 2.0 it fails linking the QtWebKit framework itself.  The problem is a mixing-and-matching the use of "hide_symbols."

The fix is the attached changes to qtwebkit 2.0 and 2.1, which add a check for "reduce_exports" in QT_CONFIG to the relevant .pro files.

[Benjamin Reed]

Created attachment 93755 [details]
patch against the 2.1 and 2.2/HEAD branches

[Zeno Albisser]

Created attachment 113143 [details]
patch for review.

[Simon Hausmann]

Comment on attachment 113143 [details]
patch for review.

r=me

Looks like it got lost in the build system refactorings. This should also be done in WebKit2.pro, but I understand that this single patch is easier to backport first :)

[Zeno Albisser]

(In reply to comment #3)
> (From update of attachment 113143 [details])
> r=me
> 
> Looks like it got lost in the build system refactorings. This should also be done in WebKit2.pro, but I understand that this single patch is easier to backport first :)

i created a separate bugreport for that, and will fix it asap: https://bugs.webkit.org/show_bug.cgi?id=71276

[WebKit Review Bot]

Comment on attachment 113143 [details]
patch for review.

Clearing flags on attachment: 113143

Committed r98947: <http://trac.webkit.org/changeset/98947>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.