/root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/include/mozilla/Casting.h:42:1: error: mangled name for 'void mozilla::BitwiseCast(From, To*) [with To = void*; From = float (*)(float) throw ()]' will change in C++17 because the exception specification is part of a function type [-Werror=c++1z-compat] BitwiseCast(const From aFrom, To* aResult) ^~~~~~~~~~~ /root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/include/mozilla/Casting.h:57:1: error: mangled name for 'To mozilla::BitwiseCast(From) [with To = void*; From = float (*)(float) throw ()]' will change in C++17 because the exception specification is part of a function type [-Werror=c++1z-compat] BitwiseCast(const From aFrom) ^~~~~~~~~~~ This is breaking the build with --enable-warnings-as-errors is used

Is this literally an error in the header, or is it an error in some *use* of BitwiseCast? The message suggests to me that it's the latter, in a use of BitwiseCast that explicitly parametrizes a BitwiseCast call with the From type. If that's what it is, simply adding in |throw| to the explicitly-specified From (with whatever macro-guard is needed for backwards compatibility) should do the trick. But it's impossible to say without more error context. Mind providing it?

Of course, sorry about that: /usr/bin/g++-7 -std=gnu++11 -o Unified_cpp_js_src10.o -c -I/root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/system_wrappers -include /root/firefox-gcc-last/config/gcc_hidden.h -DDEBUG=1 -DTRACING=1 -DENABLE_BINARYDATA -DENABLE_SIMD -DENABLE_SHARED_ARRAY_BUFFER -DEXPORT_JS_API -DJS_HAS_CTYPES '-DDLL_PREFIX="lib"' '-DDLL_SUFFIX=".so"' -DFFI_BUILDING -DMOZ_HAS_MOZGLUE -I/root/firefox-gcc-last/js/src -I/root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/js/src -I/root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/js/src/ctypes/libffi/include -I/root/firefox-gcc-last/js/src/ctypes/libffi/src/x86 -I/root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/include -I/root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/include/nspr -fPIC -DMOZILLA_CLIENT -include /root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/js/src/js-confdefs.h -MD -MP -MF .deps/Unified_cpp_js_src10.o.pp -Wall -Wc++11-compat -Wempty-body -Wignored-qualifiers -Woverloaded-virtual -Wpointer-arith -Wsign-compare -Wtype-limits -Wunreachable-code -Wwrite-strings -Wno-invalid-offsetof -Wc++14-compat -Wc++1z-compat -Wno-error=maybe-uninitialized -Wno-error=deprecated-declarations -Wno-error=array-bounds -fno-rtti -fno-exceptions -fno-math-errno -pthread -pipe -g -freorder-blocks -O3 -fno-omit-frame-pointer -Werror -Wno-shadow -Werror=format -fdiagnostics-color /root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/js/src/Unified_cpp_js_src10.cpp In file included from /root/firefox-gcc-last/js/src/jstypes.h:25:0, from /root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/include/js/Utility.h:27, from /root/firefox-gcc-last/js/src/jsalloc.h:18, from /root/firefox-gcc-last/js/src/jsatom.h:12, from /root/firefox-gcc-last/js/src/jsscript.h:18, from /root/firefox-gcc-last/js/src/jit/BytecodeAnalysis.h:10, from /root/firefox-gcc-last/js/src/jit/BytecodeAnalysis.cpp:7, from /root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/js/src/Unified_cpp_js_src10.cpp:2: /root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/include/mozilla/Casting.h:42:1: error: mangled name for 'void mozilla::BitwiseCast(From, To*) [with To = void*; From = float (*)(float) throw ()]' will change in C++17 because the exception specification is part of a function type [-Werror=c++1z-compat] BitwiseCast(const From aFrom, To* aResult) ^~~~~~~~~~~ /root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/include/mozilla/Casting.h:57:1: error: mangled name for 'To mozilla::BitwiseCast(From) [with To = void*; From = float (*)(float) throw ()]' will change in C++17 because the exception specification is part of a function type [-Werror=c++1z-compat] BitwiseCast(const From aFrom)

Hmm, that doesn't identify the triggering code. :-( There aren't too many BitwiseCast calls in js/, tho. Skimming them, I don't see any that *explicitly* pass in float(*)(float) as From-type. So my guess is that this is what's responsible: /home/jwalden/moz/slots/js/src/jstypes.h 208:#define JS_FUNC_TO_DATA_PTR(type, fun) (mozilla::BitwiseCast<type>(fun)) 209:#define JS_DATA_TO_FUNC_PTR(type, ptr) (mozilla::BitwiseCast<type>(ptr)) and if we inserted in second template parameters of |decltype(&fun)| and |decltype(ptr)| that might be a fix. But it's hard to say without testing locally. Maybe I need to spin up a new gcc build again...

gcc renamed the warning to noexcept-type: /root/firefox-gcc-last/obj-x86_64-pc-linux-gnu/dist/include/mozilla/Casting.h:42:1: error: mangled name for 'void mozilla::BitwiseCast(From, To*) [with To = void*; From = float (*)(float) throw ()]' will change in C++17 because the exception specification is part of a function type [-Werror=noexcept-type]

As I don't know how to fix it (and it is only when we will switch to C++17), I am proposing to disable the failure when we encounter this warning.

It seems like this should be fixed by replacing throw() everywhere it's used with a macro that expands to either throw() or noexcept(true) depending on the value of __cplusplus, similar to bug 1383919.

Comment on attachment 8893783 [details] bug 1320656 - When exists, add -Wno-noexcept-type to the js build flags until we have a proper fix https://reviewboard.mozilla.org/r/164882/#review172130 ::: js/src/moz.build:737 (Diff revision 1) > > if CONFIG['GNU_CXX']: > # Also disable strict-aliasing for GCC compiler, that is enabled by default > # starting with version 7.1, see Bug 1363009 > - CXXFLAGS += ['-Wno-shadow', '-Werror=format', '-fno-strict-aliasing'] > + CXXFLAGS += ['-Wno-shadow', '-Werror=format', '-fno-strict-aliasing', > + '-Wno-error=noexcept-type'] Add a comment pointing to this bug. Also, a note about what the warning is would help relativize. Per https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html: "Warn if the C++1z feature making noexcept part of a function type changes the mangled name of a symbol relative to C++14. Enabled by -Wabi and -Wc++1z-compat." Unless that's exposed as a public ABI, this is pretty much "so what?"

Pushed by sledru@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/61cf451831f5 Disable the failure on -Wnoexcept-type until we have a proper fix r=glandium

Backed out for bustage in RegExp.o: https://hg.mozilla.org/integration/autoland/rev/03bdb029f9c6a82c7bbc25c25ebdc87e240d8f94 Push with bustage: https://treeherder.mozilla.org/#/jobs?repo=autoland&revision=61cf451831f5673de926d67d74d9e03f5cbb642d&filter-resultStatus=testfailed&filter-resultStatus=busted&filter-resultStatus=exception&filter-resultStatus=retry&filter-resultStatus=runnable Build log: https://treeherder.mozilla.org/logviewer.html#?job_id=122213998&repo=autoland

Fails because of: [task 2017-08-10T07:59:39.784045Z] 07:59:39 INFO - cc1plus: error: -Werror=noexcept-type: no option -Wnoexcept-type I guess replacing it by -Wno-noexcept-type could work...

I bet not. This is likely a new flag that is not supported by GCC 4.9.

As you expected, cc1plus: error: unrecognized command line option "-Wno-noexcept-type" [-Werror] I will add the detection of the flag.

Pushed by sledru@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/67963f1a665a When exists, add -Wno-noexcept-type to the js build flags until we have a proper fix r=glandium

The leave-open keyword is there and there is no activity for 6 months. :froydnj, maybe it's time to close this bug?

I guess we've fixed this, and we can open a new bug for the proper fix referenced in comment 21? (Also, it amuses me to redirect sylvestre's bot inquires to sylvestre ;)

Yeah, Irony, and I reported it! :p