Closed Bug 686280 Opened 14 years ago Closed 14 years ago

build error in xpcwrappedjsclass.cpp

Categories

(Core :: JavaScript Engine, defect)

x86_64
Linux
defect
Not set
normal

Tracking

()

RESOLVED FIXED
mozilla9
Tracking Status
firefox8 --- affected
firefox9 --- affected

People

(Reporter: arno, Unassigned)

References

Details

Attachments

(1 file)

Hi, I currently cannot build mozilla anymore. I get following error: In file included from /home/arno/mozilla/js/src/jsobj.h:59:0, from /home/arno/mozilla/js/src/xpconnect/src/xpcpublic.h:45, from /home/arno/mozilla/js/src/xpconnect/src/xpcprivate.h:53, from /home/arno/mozilla/js/src/xpconnect/src/xpcwrappedjsclass.cpp:45: /home/arno/mozilla/js/src/jsvector.h: In static member function ‘static void js::VectorImpl<T, N, AP, IsPod>::moveConstruct(T*, const U*, const U*) [with U = js::Value, T = js::Value, long unsigned int N = 8ul, AP = js::TempAllocPolicy, bool IsPod = false]’: /home/arno/mozilla/js/src/jsvector.h:620:5: instantiated from ‘bool js::Vector<T, MinInlineCapacity, AllocPolicy>::convertToHeapStorage(size_t) [with T = js::Value, long unsigned int N = 8ul, AllocPolicy = js::TempAllocPolicy, size_t = long unsigned int]’ /home/arno/mozilla/js/src/jsvector.h:637:34: instantiated from ‘bool js::Vector<T, MinInlineCapacity, AllocPolicy>::growStorageBy(size_t) [with T = js::Value, long unsigned int N = 8ul, AllocPolicy = js::TempAllocPolicy, size_t = long unsigned int]’ /home/arno/mozilla/js/src/jsvector.h:673:5: instantiated from ‘bool js::Vector<T, MinInlineCapacity, AllocPolicy>::growByImpl(size_t) [with bool InitNewElems = false, T = js::Value, long unsigned int N = 8ul, AllocPolicy = js::TempAllocPolicy, size_t = long unsigned int]’ /home/arno/mozilla/js/src/jsvector.h:699:34: instantiated from ‘bool js::Vector<T, MinInlineCapacity, AllocPolicy>::growByUninitialized(size_t) [with T = js::Value, long unsigned int N = 8ul, AllocPolicy = js::TempAllocPolicy, size_t = long unsigned int]’ /home/arno/mozilla/js/src/jscntxt.h:2432:9: instantiated from ‘bool js::AutoVectorRooter<T>::resize(size_t) [with T = js::Value, size_t = long unsigned int]’ /home/arno/mozilla/js/src/xpconnect/src/xpcwrappedjsclass.cpp:1462:26: instantiated from here /home/arno/mozilla/js/src/jsvector.h:93:13: error: call of overloaded ‘Value(js::MoveRef<const js::Value>)’ is ambiguous /home/arno/mozilla/js/src/jsvector.h:93:13: note: candidates are: /home/arno/mozilla/js/src/jsvalue.h:334:7: note: constexpr js::Value::Value(const js::Value&) /home/arno/mozilla/js/src/jsvalue.h:334:7: note: constexpr js::Value::Value(js::Value&&) make[6]: *** [xpcwrappedjsclass.o] Erreur 1 I'm using gcc-4.6 Here is my .mozconfig: mk_add_options MOZ_OBJDIR=@TOPSRCDIR@/obj-builddir mk_add_options MOZ_BUILD_PROJECTS="xulrunner browser" mk_add_options MOZ_CO_PROJECT="xulrunner browser" ac_add_app_options xulrunner --enable-application=xulrunner ac_add_app_options browser --enable-application=browser mk_add_options MOZ_MAKE_FLAGS="-j2" ac_add_options --disable-crashreporter ac_add_options --disable-javaxpcom ac_add_options --enable-chrome-format=flat ac_add_options --with-system-zlib ac_add_options --with-system-jpeg ac_add_options --with-system-bz2 ac_add_options --with-system-sqlite ac_add_options --enable-system-cairo ac_add_options --disable-debug-symbols
I am getting this too. It looks more of a JS core issue than an XPConnect issue to me. And probably a case of gcc 4.6 being pickier than previous versions.
Assignee: nobody → general
Component: XPConnect → JavaScript Engine
QA Contact: xpconnect → general
It looks like moveRef and moveConstructor was introduced in bug 672728
Blocks: 672728
Thanks Luke, for being the unholy master of C++ and writing the whole patch on my laptop. I'll let you explain what is actually going wrong here.
According to the error message, the problem is that are trying to construct a js::Value given a const js::Value & and there is an ambiguity between Value(const Value &) and Value(Value &&) (yes, that is the C++11 "rval reference"). This is insane so I assume something in GCC is broken, either the check or the error message. Anyhow, the patch simply fixes the fact that if you "move" from a temporary (of type const T &), you should get a MoveRef<T>, not MoveRef<const T> (otherwise, how could you steal the temporary's guts?).
Attachment #560394 - Flags: review?(jimb)
Without -std=gnu++0x, the error becomes: In file included from /mnt/js/src/jsobj.h:59:0, from /mnt/js/src/xpconnect/src/xpcpublic.h:45, from /mnt/js/src/xpconnect/src/xpcprivate.h:53, from /mnt/js/src/xpconnect/src/xpcwrappedjsclass.cpp:45: /mnt/js/src/jsvector.h: In static member function ‘static void js::VectorImpl<T, N, AP, IsPod>::moveConstruct(T*, const U*, const U*) [with U = js::Value, T = js::Value, long unsigned int N = 8ul, AP = js::TempAllocPolicy, bool IsPod = false]’: /mnt/js/src/jsvector.h:620:5: instantiated from ‘bool js::Vector<T, MinInlineCapacity, AllocPolicy>::convertToHeapStorage(size_t) [with T = js::Value, long unsigned int N = 8ul, AllocPolicy = js::TempAllocPolicy, size_t = long unsigned int]’ /mnt/js/src/jsvector.h:637:34: instantiated from ‘bool js::Vector<T, MinInlineCapacity, AllocPolicy>::growStorageBy(size_t) [with T = js::Value, long unsigned int N = 8ul, AllocPolicy = js::TempAllocPolicy, size_t = long unsigned int]’ /mnt/js/src/jsvector.h:673:5: instantiated from ‘bool js::Vector<T, MinInlineCapacity, AllocPolicy>::growByImpl(size_t) [with bool InitNewElems = false, T = js::Value, long unsigned int N = 8ul, AllocPolicy = js::TempAllocPolicy, size_t = long unsigned int]’ /mnt/js/src/jsvector.h:699:34: instantiated from ‘bool js::Vector<T, MinInlineCapacity, AllocPolicy>::growByUninitialized(size_t) [with T = js::Value, long unsigned int N = 8ul, AllocPolicy = js::TempAllocPolicy, size_t = long unsigned int]’ /mnt/js/src/jscntxt.h:2432:9: instantiated from ‘bool js::AutoVectorRooter<T>::resize(size_t) [with T = js::Value, size_t = long unsigned int]’ /mnt/js/src/xpconnect/src/xpcwrappedjsclass.cpp:1462:26: instantiated from here /mnt/js/src/jsvector.h:93:13: error: invalid cast from type ‘js::MoveRef<const js::Value>’ to type ‘const js::Value*’
Is comment 6 with the patch applied?
No, it was just for reference, as the error makes more sense there. Sorry for the confusion. The patch does solve the problem for me.
Attachment #560394 - Flags: review?(jimb) → review+
Minimized test case: template <typename T> struct MoveRef { operator T& () {} }; template <typename T> MoveRef <T> Move(T&) {} struct Thing {}; Thing foo(const Thing* p) { return Thing(Move(*p)); } Filed http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50442 .
Whiteboard: [inbound]
Target Milestone: --- → mozilla9
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
Whiteboard: [inbound]
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: