[PATCH] D69764: [clang-format] Add East/West Const fixer capability

[Florin Iucha via Phabricator via cfe-commits]

0x8000-0000 added a comment.

In D69764#2532952 <https://reviews.llvm.org/D69764#2532952>, @sammccall wrote:

> In D69764#2532666 <https://reviews.llvm.org/D69764#2532666>, @MyDeveloperDay wrote:
>
>>> What can be done to move this change along?
>>
>> I feel there has to be a fundamental acceptance that it is ok for clang-format to alter code (something it already does with sorting of includes, namespace comments).
>>
>> There were fairly strong opinions that clang-format isn't the best tool to do this (which actually I don't agree with, I think it is, as long as those capabilities are off by default and there is an acceptance they won't be perfect especially in the presence of macros due to lack of AST)
>>
>> My only thought about building another tool would be if it was a drop in replacement for clang-format (tooling allows setting of a path), but it would need to inherit all of clang-format.
>> but to me, this just feels like extra grunt work just to work around why some community don't like it.
>
> Yeah, this seems like adding a flag with extra steps.
>
> clang-format's brand is:
>
> - fast
> - semantic no-op
> - applies a consistent, project-specific style
>
> I think putting it (permanently) behind a flag or alternate binary would cut against #3. I don't like that.

It can naturally be behind a flag, with is: east, west, leave_it_be. Skittish people/teams can leave_it_be for a release or two.

> If it's buggy, this feature risks cutting against #2 (more than usual). So code supporting this feature is more critical than it was previously, and that might be a lot of heuristics.

How is the risk of this bugginess greater than the risk of any other transformation? (header sorting can expose invalid dependencies - and break the build).


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69764/new/

https://reviews.llvm.org/D69764