Failure to build 1.0.8744

[ArchangeGabriel]

Hi there,

Thanks to all your hard work lately, I’m getting close to compiling against LLVM12. I’ve having some issues still with the latest release, but not LLVM related. One I’m going to make a PR for, that is LLD-building related.

The other one is this:

In function ‘llvm::Type* getIntegerEquivalent(llvm::Type*)’,
    inlined from ‘llvm::Type* getIntegerEquivalent(llvm::Type*)’ at /build/intel-graphics-compiler/src/intel-graphics-compiler-igc-1.0.8744/IGC/Compiler/Optimizer/OpenCLPasses/JointMatrixFuncsResolutionPass.cpp:201:14:
/build/intel-graphics-compiler/src/intel-graphics-compiler-igc-1.0.8744/IGC/Compiler/Optimizer/OpenCLPasses/JointMatrixFuncsResolutionPass.cpp:213:51: error: ‘this’ pointer is null [-Werror=nonnull]
  213 |         unsigned size = matTy->getScalarSizeInBits();
      |                         ~~~~~~~~~~~~~~~~~~~~~~~~~~^~
In file included from /usr/include/llvm/IR/DerivedTypes.h:23,
                 from /usr/include/llvm/IR/DataLayout.h:26,
                 from /usr/include/llvm/IR/Module.h:25,
                 from /build/intel-graphics-compiler/src/intel-graphics-compiler-igc-1.0.8744/IGC/common/igc_resourceDimTypes.h:12,
                 from /build/intel-graphics-compiler/src/intel-graphics-compiler-igc-1.0.8744/IGC/Compiler/CodeGenPublic.h:18,
                 from /build/intel-graphics-compiler/src/intel-graphics-compiler-igc-1.0.8744/IGC/Compiler/Optimizer/OpenCLPasses/JointMatrixFuncsResolutionPass.h:11,
                 from /build/intel-graphics-compiler/src/intel-graphics-compiler-igc-1.0.8744/IGC/Compiler/Optimizer/OpenCLPasses/JointMatrixFuncsResolutionPass.cpp:9:
/usr/include/llvm/IR/Type.h: In function ‘llvm::Type* getIntegerEquivalent(llvm::Type*)’:
/usr/include/llvm/IR/Type.h:304:12: note: in a call to non-static member function ‘unsigned int llvm::Type::getScalarSizeInBits() const’
  304 |   unsigned getScalarSizeInBits() const LLVM_READONLY;
      |            ^~~~~~~~~~~~~~~~~~~

Of course I can add -Wno-error=nonnull, and I will likely do so, but figured out I should still report. ;)

[mbelicki]

Hi Bruno, thanks for reporting the issue. I will look into the other one and try to fix it on our end.

[mbelicki]

@ArchangeGabriel I've committed a fix for the issue last week (commit a1240fa). Could you verify that you are no longer observing the problem with null this pointer?

[tjaalton]

I can confirm that the fix works, just hit this build error on Debian and with a1240fa the failure is gone

[pszymich]

Great! Thanks @mbelicki and @tjaalton for verification. @ArchangeGabriel please reopen if the fix does not apply to your issue.

[ArchangeGabriel]

Sorry for the delay, can confirm that this fixes the issue.