Re: [RFC PATCH 3/4] .clang-format: do not enforce a ColumnLimit
- From
Ævar Arnfjörð Bjarmason <avarab@gmail.com>
- Date
- Jul 12, 2022, 07:03 UTC
- Message-ID
- <220712.868roy6c0s.gmgdl@evledraar.gmail.com>
- In-Reply-To
- <xmqqedyrmiu8.fsf@gitster.g>
On Mon, Jul 11 2022, Junio C Hamano wrote:
Show 16 quoted lines
> Ævar Arnfjörð Bjarmason <avarab@gmail.com> writes: > >> As with the preceding change what this leaves us with an unresolved >> question though, should we have some stricter version of "make >> style-all" that incorporates "ColumnLimit: 80", or perhaps apply it >> only on "make style", but then what if someone modifies code that >> happens to e.g. search/replace a line running afoul of the limit? > > A more important thing to think about is that there is no single > good cut-off point. When we say "wrap your lines at around 80 > columns", we mean that when there is a good place to fold at around > column 65 and the next good place is at column 82, then it is OK to > go slightly over 80 and wrap at 82, which may be better than > wrapping at 65. If the last good place to wrap is at column 72 and > the long function call at the end of the line makes you go past the > 82nd column, wrapping at column 72 might be better.
There's the story of the sufficiently smart compiler, and now the sufficiently smart formatter :)
The proposed solution here is to punt on it, which I think makes sense if we're trying to push forward clang-format.
(Which I'm really not, this RFC is something I thought I'd send in response to brian's proposal, as I'd poked at this locally a while ago, after wondering if I could make use of it myself, and whether our .clang-format was misconfigured[1]).
> I wonder if > there is an automated formatter that understands this kind of shades > of gray and lets us express that.
I don't think so, and setting the configuration to "0" is only a stop-gap, after all we'd still like it to wrap e.g. lines of a length of 150 or whatever, if it finds them somewhere.
1. I think after I found that some odd styling from one of Han-Wen's patches was the result of running clang-format, cf.: https://lore.kernel.org/git/CAFQ2z_PAqW+RS2Znaf2wwOJfdNfkjP1VV84=xaPu_1EAuX+u5w@mail.gmail.com/