On Fri, Nov 04 2022, Jeff King wrote: > On Fri, Nov 04, 2022 at 12:36:12PM +0100, Ævar Arnfjörð Bjarmason wrote: > >> > It feels like the invocation of bisect--helper ought to be passing "--" >> > itself to indicate the end of options, like: >> >> Normally you'd be righ, but there's a much easier way to do it in this >> case: cherry-pick >> https://lore.kernel.org/git/05262b6a7d1b20a0d2f2ca2090be284ffb8c679c.1661885419.git.gitgitgadget@xxxxxxxxx/ >> >> I.e. we're eating "--log" in particular, but could just stop >> understanding that option. > > We're not eating "--log" in particular, though. We're eating anything > that parse_options() understands. It would eat "--bisect-start", too. > That's less likely than "--log", but the fundamental problem would still > remain. Right, I mean the regression's anyone's likely to care about. I.e. that's an easy fix, I think run scripts using --bisect-* options are probably unlikely to exist. But yes, the ultimate fix is to use OPT_SUBCOMMAND(), but getting there takes more time...
