[I didn't get André's Email...] On 12.09.23 23:40, Ahmad Samir wrote: > On 13/9/23 00:11, apoenitz wrote: >> On Tue, Sep 12, 2023 at 11:33:17PM +0300, Ahmad Samir wrote: >>> A config file that is 80-90% correct is better than nothing. >> >> I disagree. >> >> The result of such a thing is that people submit patches matching the >> config >> 100%, deviating from the wanted style by 10-20%. Numbers are arbitrary >> here, but the effect is that in practice even long-term contributors >> start >> to submit "wrong" auto-formatted patches causing needless review >> roundtrips. >> > > The alternative is patches with arbitrary formatting; that's not better.
Sorry, no. If you submit a patch, you should just use whatever style you find in the file. If there's a mix, use whatever is adjacent to where you're coding. It's that simple. That shouldn't be too much to ask, and I, for one, am fine going into the Gerrit editor and fixing things for the submitter if that's the only thing holding me back from a +2. It's when people insist on rewriting lines because that's how QtCreator or git-clang-format format it, that the discussions start. And a clang-format file doesn't actually help here, because we have already introduced inconsistencies in qtbase and whatever we select in .clang-format will be wrong 50% of the time. Assuming, of course, that you can get it configured to be close enough to the prevailing style, which I have not seen proof of. If anyone thinks it's possible, there's a simple litmus test: apply your config to all of qt(base) and look at the diff: Is it just fixing some outliers, or is it rewriting the world? >> On top of that repeatedly discussion are started about adjusting the >> style to >> the tool, because "the tool can't be adjusted". > > True, but style arguments arise on their own too, tool or no tool. It only arises because the tool cannot represent what we traditionally call the Qt style. If there are disagreements on what that means, it's easy to draw upon a majority vote of Qt 4 code to break the tie, disregarding clang-format-infected files. I should also point out that the discussions a few years ago were nitpicking about Qt coding style violations, and the expected response from the submitter was to fix them. These days, everything is called into question because clang-format does it differently, ie. wrongly. And I'm kinda getting tired commenting on all these clang-format idiosyncrasies. If the tool was truly configurable, e.g. with an example file from which it would pick formatting, we wouldn't have this discussion, either. But what I read between the lines is that Google is happy that it can reproduce their style, and contributions to extend it are not welcome. > One way to stop the recurring discussions would be to reach a majority > consensus about a clang-format config file, format the whole code base > of each module in one go, then apply the formatting for each new commit, > less discussions. Is that ideal? no, because reaching a consensus about > coding style is not easy. There _is_ consensus. It's in the wiki. And in older modules not infected by the _clang-format file. Discussions arise because of .clang-format, not despite it. Afaict, there never was a discussion about how faithful the _clang-format represents the Qt style before it was added. If there was _any_ attempt at the above-mentioned litmus test, the template<> issue and others would have been detected immediately. > I could format it manually, probably two steps, or let the tool do it That assumes the tool can be configured to reproduce the style. Until proven, you have to do the work 50% of the time, anyway. Tool won't help. Believe me, I wouldn't like anything better than a clang-format config that shuts these discussions up. But not at the cost of rewriting 50% of our code, a large percentage of which still dates back unchanged to the beginning of the public history. Thanks, Marc -- Marc Mutz <marc.m...@qt.io> Principal Software Engineer The Qt Company Erich-Thilo-Str. 10 12489 Berlin, Germany www.qt.io Geschäftsführer: Mika Pälsi, Juha Varelius, Jouni Lintunen Sitz der Gesellschaft: Berlin, Registergericht: Amtsgericht Charlottenburg, HRB 144331 B -- Development mailing list Development@qt-project.org https://lists.qt-project.org/listinfo/development
