Use tensor_shape_to_c_string for error in check_mask_indices

[swolchok]

Rolling out for #7902

[swolchok]

Stack from ghstack (oldest at bottom):

• roll out s/IF_FALSE/UNLESS/ for ET_LOG macros #8318
• s/IF_FALSE/UNLESS/ in ET_LOG macro names #8317
• Use tensor_shape_to_c_string for error in check_glu_args #8316
• -> Use tensor_shape_to_c_string for error in check_mask_indices #8314

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/8314

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 1a17c4d with merge base 9746ce7 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[dbort]

Same here, I think it'd be less surprising if the ifdef covered the entire scope of the vars that are defined inside it.

Suggested change

	#ifdef ET_LOG_ENABLED
	auto mask_shape = executorch::runtime::tensor_shape_to_c_string(
	executorch::runtime::Span<const Tensor::SizesType>(
	index.sizes().data(), index.sizes().size()));
	auto input_shape = executorch::runtime::tensor_shape_to_c_string(
	executorch::runtime::Span<const Tensor::SizesType>(
	in.sizes().data() + in_i, index.sizes().size()));
	#endif // ET_LOG_ENABLED
	ET_LOG(
	Error,
	"The shape of mask index %s must match the sizes of the corresponding input dimensions %s.",
	mask_shape.data(),
	input_shape.data());
	#ifdef ET_LOG_ENABLED
	auto mask_shape = executorch::runtime::tensor_shape_to_c_string(
	executorch::runtime::Span<const Tensor::SizesType>(
	index.sizes().data(), index.sizes().size()));
	auto input_shape = executorch::runtime::tensor_shape_to_c_string(
	executorch::runtime::Span<const Tensor::SizesType>(
	in.sizes().data() + in_i, index.sizes().size()));
	ET_LOG(
	Error,
	"The shape of mask index %s must match the sizes of the corresponding input dimensions %s.",
	mask_shape.data(),
	input_shape.data());
	#endif // ET_LOG_ENABLED

[cccclai]

It seems like this PR breaks multiple tests

[swolchok]

It seems like this PR breaks multiple tests

ugh, why do we have tests that only run on trunk, grumble grumble

[swolchok]

suspiciously it is mostly qnn stuff + test-binary-size-linux. investigating

[swolchok]

#8403 should fix test-binary-size-linux

[swolchok]

#8404 might fix QNN, running ciflow/trunk on the stack (EDIT: looks green)