[Mlir-commits] [mlir] [Quick-fix] Fix GCC build of previous patch. (PR #111869)

[Renato Golin]

rengolin wrote:

https://lab.llvm.org/buildbot/#/builders/117/builds/2697

Compiler: GCC 11.4

```
/home/uweigand/sandbox/buildbot/mlir-s390x-linux/llvm-project/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp: In function 'llvm::SmallVector<mlir::AffineMap> getDefaultIndexingMapsForMatmul(mlir::MLIRContext*)':
/home/uweigand/sandbox/buildbot/mlir-s390x-linux/llvm-project/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp:168:10: error: could not convert 'indexingMaps' from 'SmallVector<[...],3>' to 'SmallVector<[...],6>'
  168 |   return indexingMaps;
      |          ^~~~~~~~~~~~
      |          |
      |          SmallVector<[...],3>
```

https://llvm.org/docs/ProgrammersManual.html#llvm-adt-smallvector-h

I'm not sure where the `6` is coming from, to be honest.

Perhaps a better fix is to change the return type to `SmallVector<AffineMap, 3>` or change the declaration to `SmallVector<AffineMap>`. But I'm not sure GCC will like either of them, since Clang is perfectly fine with it.

I'll try to build with GCC localy.

https://github.com/llvm/llvm-project/pull/111869