GSOC 2021 : Improvements to clang-format and clang-tidy to format code according to haiku coding guidelines | Haiku Project

Blog
1

I am Saloni B.Tech.(3rd year), Computer Science and Engineering(CSE) student of Guru Nanak Dev Engineering College, India. I have been selected for Google Summer of Code 2021 to work with Haiku on the project Improvements to clang-format and clang-tidy to format code according to haiku coding guidelines. My mentors are Preetpal Kaur and Adrien Destugues.

This is a companion discussion topic for the original entry at https://www.haiku-os.org/blog/saloni/2021-05-20_gsoc_2021__improvements_to_clang-format_and_clang-tidy_to_format_code_according_to_haiku_coding_guidelines/
10 Likes
2

Congrats! I have just reviewed some of your work and would like to offer a few tips:

  • You should rebase your code to my latest (6+ months old though) haiku-format as most of the work you have been doing had already been done there. (See #18, #21, and #22.)

  • Do not add Haiku-specific style options like I initially did here. Instead, see the much simpler approach I later took.

  • Take a look at some of my previous work for clang-format to get a feeling for how to fix bugs in, and add features to, clang-format. Pay attention to the test cases added to the test suite as it’s very easy to trigger a regression.

Good luck!

2 Likes
3

Hi!

We know of your work but some of it was removed from the current branch because it was not working correctly. For example the code to remove unneeded braces on single-statement if/else would sometime remove braces on multi-statement ones. That’s why it has been replaced by a different implementation that was extracted from a work in progress change in llvm bugtracker (and I think we can use the final version of that when it will be merged on llvm side)

4

For example the code to remove unneeded braces on single-statement if/else would sometime remove braces on multi-statement ones.

I already removed it in this commit last year based on feedback from korli and then you. See the rationale here.

I also addressed all other issues (except #17) reported by you in subsequent commits. So as of late last year, there were no open issues (other than #17) that warranted further investigation.

That’s why it has been replaced by a different implementation that was extracted from a work in progress change in llvm bugtracker (and I think we can use the final version of that when it will be merged on llvm side)

Are you talking about https://reviews.llvm.org/D95168? But it’s kind of the opposite of removing redundant braces. Anyway, beware that it didn’t get landed after all because it failed when tested against a large code base.