AaronBallman wrote: > I thought the demo was pretty convincing 😅 > > The motivation is to have less scrollback.
But it's not on by default, so everyone will have all that scrollback anyway unless they're willing to fiddle with their build system to add the flag, which requires them to know about the flag in the first place. I'm not convinced this will be used in practice very often. > I suppose this is also useful for filling up agents's context windows less > quickly. Not a motivating reason for the extra maintenance burden of having to support the flag forever. > But IMHO it's nice for humans too. I try to keep Chromium's build output > clean (-Werror everywhere, build steps are only allowed to print when they > fail). But clang is pretty wordy when it fails, and I thought I might try to > do something about it 🙂 > > (-fno-caret-diagnostics is too aggressive; I think seeing details about the > actual diagnostic is very useful, while snippets for notes are much less > useful.) > > Does that make sense? Want me to put some of that in the commit description? It does, thank you for the extra explanation! I definitely appreciate trying to improve Clang's diagnostic output, and template stacks (and macro expansion stacks) can get chatty. We do have `-fdiagnostics-show-note-include-stack` which is in a similar vein but the opposite direction (shows MORE information instead of LESS). Same for `-fdiagnostics-show-inlining-chain`. I'm not seeing other times where we default to showing more information but let the user show less via flags, so I mostly worry about whether the feature will be enabled enough to justify having it in-tree. For example, those two flags I mentioned are not well-used in practice; searching for them on sourcegraph shows about 20 uses of `-fdiagnostics-show-note-include-stack` and about 10 uses of `-fdiagnostics-show-inlining-chain` between Makefiles and CMake. https://github.com/llvm/llvm-project/pull/216250 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
