On Sun, Mar 8, 2009 at 12:48 AM, Jeff King <peff@xxxxxxxx> wrote: > On Sat, Mar 07, 2009 at 11:07:46PM +0200, Felipe Contreras wrote: > >> On Sat, Mar 7, 2009 at 7:14 PM, Jeff King <peff@xxxxxxxx> wrote: >> > The conversion to parse_opt left this as NULL; on glibc >> > systems, the usage message prints >> > >> > --bool-or-int (null) >> > >> > and on other ones, segfaults. >> >> Shouldn't then OPT_BIT make sure there is no crash? > > Perhaps, but it doesn't (and I assume you mean usage_with_help, as > OPT_BIT is just filling in the struct). It's not clear what a NULL help > parameter should do, though. Hide the option? Show no help description? > There are already ways to accomplish both of those. Yeah, I meant usage_with_help. I don't know what should be done, but I think two things should be achieved: a) don't crash b) encourage the options to always have a description Perhaps not showing the option at all, or perhaps showing "**EMPTY**". >> I was surprised when it didn't complain. I thought on making it "" but >> I wanted to make it visible that there was no documentation for that, >> which is the reason I left it that way. > > OK. I think there are really valid options: > > 1. it's there with a description (which is what my patch does) > > 2. it's there without a description, because it's obvious what it does > coming after --bool and --int I don't think it's obvious, that partly why I didn't fill the description. > 3. it's hidden > > I really don't care which. But what is there now is broken. Definitely, your patch must be applied ASAP. Minor nitpick: "value is interpreted either as bool or int" The value is what it is, the --boo-or-int option doesn't change the value, just how it is interpreted. -- Felipe Contreras -- 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
