1027815 - access violation when using JS::CompileOptions in opt build

Status: RESOLVED DUPLICATE of bug 1083464

(Core :: JavaScript Engine, defect)

Description

[Franck]

I configured JS with the following options:
  (apply patch bug-975011-patch (configure, build nsprpub for js/src standalone builds))
  js/src/configure --enable-optimize=-Ox --enable-strip --with-windows-version=601 --enable-threadsafe --disable-gcgenerational --enable-exact-rooting --without-intl-api --enable-ctypes --disable-tests


Then I run the following code:
  int main( int argc, char* argv[] ) {

	JS_Init();
	JSRuntime *rt = JS_NewRuntime( 32L * 1024L * 1024L );
	JSContext *cx = JS_NewContext( rt, 8192 );
	JS_BeginRequest( cx );
	
	JS::CompileOptions co( cx );

	JS_EndRequest( cx );
	JS_DestroyContext( cx );
	JS_DestroyRuntime( rt );
	JS_ShutDown();
	return 0;
  }


The msvc2013 runtime error is:
  A buffer overrun has occurred in foo.exe which has corrupted the program's internal state.

Comment 1

[Boris Zbarsky [:bzbarsky]]

The code that you ran... was it compiled with the same set of ifdefs as spidermonkey itself?

Comment 2

[Franck]

not sure, I just #include <jsapi.h> and use the default ones: WIN32 NDEBUG _CONSOLE _LIB

Comment 3

[Boris Zbarsky [:bzbarsky]]

OK.  In a debugger, where exactly are you crashing?

Comment 4

[Franck]

The raw message is: 
  Unhandled exception at 0x752BCB49 in foo.exe: 0xC0000005: Access violation executing location 0x00000000.

The source code is unavailable (after the return 0;)

Comment 5

[Franck]

Note that the issue is solved by disabling "Whole Program Optimization" of my embedding.

Comment 6

[Philip Chimento [:ptomato]]

I had a similar problem, but it was indeed caused by not having the same set of ifdefs; I had -DJS_DEBUG and -DJSGC_USE_EXACT_ROOTING, but not -DDEBUG. I think that's probably what's going on above.

Comment 7

[Philip Chimento [:ptomato]]

I've come to think this is actually a duplicate of https://bugzilla.mozilla.org/show_bug.cgi?id=1083464