On 12/01/10 15:31, Jonathan Nieder wrote: > The PARSE_OPT_LITERAL_ARGHELP flag allows a program to override the > standard "<argument> for mandatory, [argument] for optional" markup in > its help message. Extend it to override the usual "no text for > disallowed", too (for the PARSE_OPT_NOARG | PARSE_OPT_LITERAL_ARGHELP > case, which was previously meaningless), to be more intuitive. > > The motivation is to allow update-index to correctly advertise > > --cacheinfo <mode> <object> <path> > add the specified entry to the index > > while abusing PARSE_OPT_NOARG to disallow the "sticked form" > > --cacheinfo=<mode> <object> <path> > > Noticed-by: Stephen Boyd <bebarino@xxxxxxxxx> > Signed-off-by: Jonathan Nieder <jrnieder@xxxxxxxxx> > --- > This seems like the intuitive thing to do, but the motivating use case > is iffy. Might be better to introduce a PARSE_OPT_NOSTICKED flag. parse-options should accept both forms of --cacheinfo above if the option isn't marked PARSE_OPT_NOARG. Marking it NOARG to get rid of the equals sign in the usage seems wrong when both the equals sign and no equals sign can be accepted. The fault lies in the implementation of the low-level callback. It needs to handle more of the parsing. If --cacheinfo is marked PARSE_OPT_NOARG it should show --cacheinfo If --cacheinfo is marked PARSE_OPT_OPTARG it should show --cacheinfo[=<argh>] If --cacheinfo isn't marked either of the above it should show --cacheinfo <argh> which looks like what you want and it also says it takes an argument unconditionally (perfect?). On top of that, the equals sign is optional (or at least should be). At the point of get_value() p->opt will contain whatever is after the equals sign for the option that is parsed. For options which take no value (NOARG) they'll fail the test if they have anything with an equals sign after it. By not marking the cacheinfo option NOARG, we've successfully gotten what we want so far. Now we just need to make the cacheinfo callback a bit more like get_arg() and have it look at the context argument. If p->opt is set, we should pull <mode> out of that and make sure we have two more arguments. If p->opt isn't set, we should make sure we have 3 arguments and basically do what is already there. I know this is a bit more code, but it also means that we don't have this approach bite someone else down the line when they make assumptions about options marked as NOARG not taking arguments. We should probably add another check like "if argh is set and PARSE_OPT_NOARG is true error out" so this can't be done. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html