Regression with clang-format 14 & 15

[OldBragon]

The line

LongClassName<AnotherLongClassName>{param}->longFunctionName(param0, param1);

is formatted to:

LongClassName<AnotherLongClassName> { param } -> longFunctionName(param0, param1);

with Clang 13 and the attached configuration file, and formatted to:

    LongClassName<AnotherLongClassName>
    {
        param
        } -> longFunctionName(param0, param1);

with Clang 14 & 15.

Reproduced with https://clang-format-configurator.site/
Thanks
clang-format.txt

[llvmbot]

@llvm/issue-subscribers-clang-format

[rymiel]

FWIW, clang-format-16 will now format it to the following:

LongClassName<AnotherLongClassName>
{
    param
} -> longFunctionName(param0, param1);

as part of 1fa115b.
The difference between 15 and 16 is intentional, as it was a fix as part of 1fa115b. I cannot comment if the listed change is a "regression", I predict it is unlikely to be "reverted".

[HazardyKnusperkeks]

But this is not part of a requires clause, is it?

[rymiel]

That's what the commit was intended to fix, but it actually affects all braced blocks (sort of discussed in https://reviews.llvm.org/D134626). My bad, I should have probably worded the commit better