sammccall added a comment. (Sorry for arriving at this late)
At a strategic level, I have some concerns here: the fact that clang-format generally doesn't touch the token sequence isn't an accident. e.g. formatting `int x;;` will insert a newline rather than deleting the redundant semicolon. Like the one in this patch, that would be a useful feature, but it's a path the clang-format authors deliberately decided to close off. The conservative approach means that pseudo-parser gets things wrong, we get broken formatting, and ~never broken code. This means: - a simpler mental model. People are happier to blindly apply it, not to require review of formatting changes, etc - the ability to require clang-formatting as a pre-commit hook, knowing that at worst people will have to badly format their code - a lesser risk when rolling out new versions of clang-format from (our org does this from head weekly) (There are exceptions, e.g. comment splitting, but they're pretty rare) @klimek in case he has thoughts. ================ Comment at: clang/lib/Format/Format.cpp:768 LLVMStyle.CommentPragmas = "^ IWYU pragma:"; + LLVMStyle.ConstStyle = FormatStyle::CS_Leave; LLVMStyle.CompactNamespaces = false; ---------------- MyDeveloperDay wrote: > lebedev.ri wrote: > > MyDeveloperDay wrote: > > > lebedev.ri wrote: > > > > Based on code reviews, this should be `CS_West`. > > > I would tend to agree for the LLVM project itself, but I really don't > > > feel I can change this due to the huge number of projects which have > > > BasedOnStyle: LLVM (not to mention most other style inherit from LLVM > > > like Google, Chromium etc...) > > > > > > I think this needs to be an opt-in change in the .clang-format (at least > > > for now) > > There is already a number of stylistic LLVM-specific decisions in LLVM style > > * 80-col wrap > > * pointer alignment to the right > > * spaces instead of tabs > > * ??? > > I think it first and foremost is an LLVM style, and i don't > > really recall any guarantees that it must not/will not change. > Again I do agree, but I think those rules were established before > clang-format really gained such popular usage, those coming to clang-format > changed those probably straight away if they didn't agree with their > established their style. > > Changing something under everyone now feels like I'd have a different agenda > (for west const to win), and that that is not my goal. I'm simply trying to > facilitate compliance. > > I think we can agree the east/west const rebellion is a contentious issue, > (as where the previous formatting wars!), my aim here is to help try and > remove the discussion from the debate and let people move on! but I think > what it doesn't need is someone like clang-format being considered the bad > guy by dictating a default style globally. > > I did a quick search on github, the hits to "const int" and "int const" alone > are too eye-watering. I really don't want to break the internet with this. > For non-style reasons, I do think this should be CS_Leave in every built-in style, so that adjusting const would have to at least be explicitly in .clang-format. CHANGES SINCE LAST ACTION https://reviews.llvm.org/D69764/new/ https://reviews.llvm.org/D69764 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
