[clang-format] Recognize nested blocks

[sstwcw]

Summary

Now a block gets recognized as a block if it contains a block.

The special cases we added don't have additional tests because they
are already tested for example in FormatTest.FormatsLambdas and
FormatTestJS.ContainerLiterals.

Fixes https://github.com/llvm/llvm-project/issues/33891

[sstwcw]

• Remove the special case for operator[], it isn't necessary.

[galenelias]

Coincidentally I also sent out a review to fix this issue yesterday, but went with a different approach of trying to scope the ProbablyBracedList logic by just looking at the lbrace previous token.

https://reviews.llvm.org/D150403

[HazardyKnusperkeks]

On this change I like the token annotator tests, and reusing the same stack instead of adding a second one which always has the same size.
But the code changes in D150403 seem to be better, at least I couldn't think directly on something which may not work.

clang/lib/Format/UnwrappedLineParser.cpp

492

Shouldn't it be called NextBraceIsProbablyLambda (or similar)?

659

How does this work on

{
int foo[]={2,5};
int bar[]{2};
}

or

{
SomeClass foo[]={{arg1, whatever}, {arg2, dontcare}};
SomeClass foo[]{{arg1, whatever}, {arg2, dontcare}};
}

?

clang/unittests/Format/TokenAnnotatorTest.cpp

43

Nice!

[sstwcw]

I like D150403 better.