1221055 - valgrind reports many warnings when running mochitest on MacOS

Status: NEW

(Core :: Graphics, defect)

Description

[Jon Coppeard (:jonco)]

Attachment: mochilog.zip

While investigating bug 1217177 I ran the browser_syncui mochitest on MacOS under valgrind, and it reported many many warnings that appear to be related the system's graphics libraries.

I built firefox with: --enable-valgrind --disable-jemalloc --disable-replace-malloc --enable-debug --enable-optimize

I ran the test with: mach mochitest --debugger="valgrind" --debugger-args="--smc-check=all-non-file --vex-iropt-register-updates=allregs-at-mem-access --show-mismatched-frees=no --dsymutil=yes" browser_syncui.js

Some examples are:

==97458== Conditional jump or move depends on uninitialised value(s)
==97458==    at 0x1005807BF: _CFMachPortCreateWithPort2 (in /System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation)
==97458==    by 0x107A6AA76: CVDisplayLink::init() (in /System/Library/Frameworks/CoreVideo.framework/Versions/A/CoreVideo)
==97458==    by 0x107A6A798: CVCGDisplayLink::initWithCGDisplays(unsigned int*, long, int*) (in /System/Library/Frameworks/CoreVideo.framework/Versions/A/CoreVideo)
==97458==    by 0x107A6A5DE: CVDisplayLinkCreateWithCGDisplays (in /System/Library/Frameworks/CoreVideo.framework/Versions/A/CoreVideo)
==97458==    by 0x107A6C528: CVDisplayLinkCreateWithActiveCGDisplays (in /System/Library/Frameworks/CoreVideo.framework/Versions/A/CoreVideo)
==97458==    by 0x10A9299AA: OSXVsyncSource::OSXDisplay::EnableVsync() (gfxPlatformMac.cpp:510)
==97458==    by 0x10A928C3D: gfxPlatformMac::CreateHardwareVsyncSource() (gfxPlatformMac.cpp:627)
==97458==    by 0x10A923136: gfxPlatform::Init() (gfxPlatform.cpp:628)
==97458==    by 0x10A922064: gfxPlatform::GetPlatform() (gfxPlatform.cpp:455)
==97458==    by 0x10C7C648A: nsRefreshDriver::ChooseTimer() const (nsRefreshDriver.cpp:699)
==97458==    by 0x10C7C78A3: nsRefreshDriver::EnsureTimerStarted(nsRefreshDriver::EnsureTimerStartedFlags) (nsRefreshDriver.cpp:1047)
==97458==    by 0x10C7C7A2F: nsRefreshDriver::MostRecentRefresh() const (nsRefreshDriver.cpp:936)

==97458== Thread 34 Compositor:
==97458== Invalid write of size 8
==97458==    at 0x139B4FF4D: IntelStateHeaps::allocateHeap(IntelStateHeaps::HeapInfo&, unsigned int, unsigned int) (in /System/Library/Extensions/AppleIntelHD5000GraphicsGLDriver.bundle/Contents/MacOS/AppleIntelHD5000GraphicsGLDriver)
==97458==    by 0x139B4FE65: IntelStateHeaps::initHeaps(GLDShareGroupRec*) (in /System/Library/Extensions/AppleIntelHD5000GraphicsGLDriver.bundle/Contents/MacOS/AppleIntelHD5000GraphicsGLDriver)
==97458==    by 0x139B4F8DB: GenContext::Initialize(GLDContextRec*) (in /System/Library/Extensions/AppleIntelHD5000GraphicsGLDriver.bundle/Contents/MacOS/AppleIntelHD5000GraphicsGLDriver)
==97458==    by 0x139B4EFBB: Gen7Context::Initialize(GLDContextRec*) (in /System/Library/Extensions/AppleIntelHD5000GraphicsGLDriver.bundle/Contents/MacOS/AppleIntelHD5000GraphicsGLDriver)
==97458==    by 0x139B4E993: glrInitializeContext (in /System/Library/Extensions/AppleIntelHD5000GraphicsGLDriver.bundle/Contents/MacOS/AppleIntelHD5000GraphicsGLDriver)
==97458==    by 0x124988E3E: gldCreateContext (in /System/Library/PrivateFrameworks/GPUSupport.framework/Versions/A/Libraries/libGPUSupportMercury.dylib)
==97458==    by 0x13694C83D: gliCreateContextWithShared (in /System/Library/Frameworks/OpenGL.framework/Versions/A/Resources/GLEngine.bundle/GLEngine)
==97458==    by 0x103FA7BF6: CGLCreateContext (in /System/Library/Frameworks/OpenGL.framework/Versions/A/OpenGL)
==97458==    by 0x100D634F4: -[NSOpenGLContext initWithFormat:shareContext:] (in /System/Library/Frameworks/AppKit.framework/Versions/C/AppKit)
==97458==    by 0x10A80614E: mozilla::gl::GLContextProviderCGL::CreateForWindow(nsIWidget*) (GLContextProviderCGL.mm:216)
==97458==    by 0x10A8F6A19: mozilla::layers::CompositorOGL::CreateContext() (CompositorOGL.cpp:125)
==97458==    by 0x10A8F6FAF: mozilla::layers::CompositorOGL::Initialize() (CompositorOGL.cpp:220)
==97458==  Address 0x132f2e038 is not stack'd, malloc'd or (recently) free'd

==97458== Invalid write of size 8
==97458==    at 0x1057D6C75: _platform_bzero$VARIANT$Unknown (in /usr/lib/system/libsystem_platform.dylib)
==97458==    by 0x1234402E23BD: gldReadTextureData (in /System/Library/Extensions/GeForceGLDriver.bundle/Contents/MacOS/GeForceGLDriver)
==97458==    by 0x12344033107E: gldBlitFramebufferData (in /System/Library/Extensions/GeForceGLDriver.bundle/Contents/MacOS/GeForceGLDriver)
==97458==    by 0x123440331290: gldBlitFramebufferData (in /System/Library/Extensions/GeForceGLDriver.bundle/Contents/MacOS/GeForceGLDriver)
==97458==    by 0x136AF2260: gliUpdateDispatchState (in /System/Library/Frameworks/OpenGL.framework/Versions/A/Resources/GLEngine.bundle/GLEngine)
==97458==    by 0x136A664E9: gleBindFramebuffer (in /System/Library/Frameworks/OpenGL.framework/Versions/A/Resources/GLEngine.bundle/GLEngine)
==97458==    by 0x1369A1578: glBindFramebuffer_Exec (in /System/Library/Frameworks/OpenGL.framework/Versions/A/Resources/GLEngine.bundle/GLEngine)
==97458==    by 0x10A7E22B0: mozilla::gl::GLContext::InitWithPrefix(char const*, bool) (GLContext.h:1929)
==97458==    by 0x10A8061EF: mozilla::gl::GLContextProviderCGL::CreateForWindow(nsIWidget*) (GLContextProviderCGL.mm:251)
==97458==    by 0x10A8F6A19: mozilla::layers::CompositorOGL::CreateContext() (CompositorOGL.cpp:125)
==97458==    by 0x10A8F6FAF: mozilla::layers::CompositorOGL::Initialize() (CompositorOGL.cpp:220)
==97458==    by 0x10A8D7516: mozilla::layers::CompositorParent::InitializeLayerManager(nsTArray<mozilla::layers::LayersBackend> const&) (CompositorParent.cpp:1454)
==97458==  Address 0x12436b000 is not stack'd, malloc'd or (recently) free'd

Full log is attached.

Comment 1

[Julian Seward [:jseward]]

Most of those are due to Valgrind not correctly tracking which parts
of the address space are validly addressable.  That can happen with 
graphics hardware that is mapped into user space though some magic
ioctls.

I can look at this further if it is of concern.  I suspect there are
no mozilla-specific issues, although it is hard to say for sure without
looking further at it.

Comment 2

[Jon Coppeard (:jonco)]

(In reply to Julian Seward [:jseward] from comment #1)
No particular concern here, although it would be nice to be able to run Valgrind on MacOS without getting a million warnings.  Feel free to close.