On Tue, Apr 06, 2021 at 02:57:46PM -0700, Junio C Hamano wrote: > Jeff King <peff@xxxxxxxx> writes: > > > I imagine something like this: > > ... > > would work, but I didn't really test it. There's another hitch, which is > > that this subsystem has all been re-written in C. So we'd really want to > > implement it in the new code (possibly in both places, though maybe it > > is time to consider cutting over from the perl script to the C one by > > deafult?). > > > > But hopefully this illustrates the general idea, and gives somebody > > interested in the feature enough to work up their own patch. > > These "cosmetic appearance" configuration that would affect the > output from diff shown to the user would not be limited to just the > .noprefix, though. Depending on the users, they would care just as > deeply about any of these: > > .context > .interHunkContext > .mnemonicPrefix > .noprefix > .relative > .orderFile > > as Nikita does for .noprefix to send a bug report. > > Luckily or unluckily, .suppressBlankEmpty and the per-filetype > .xfuncname patterns do impact the output from the plumbing, because > git_diff_basic_config() does read them, even though they are merely > "cosmetic" configurations. > > I am unsure how much we should cater to end-user controlled > configuration when we are generating diff output for our own > consumption, but if we were to tweak "add -p" and friends to > pay attention to .noprefix, we probably should do the same for > all the others. Yes. We already have gone through this with other options (e.g., diff.algorithm). I would be happy if somebody wanted to handle the complete set. But I am also OK with stumbling towards completeness, as people who care about a particular option plumb it through. -Peff
