From: Jeff King Date: Tue, 10 Mar 2009 17:59:04 GMT Subject: Re: [PATCH 2/3] config: set help text for --bool-or-int Message-ID: <20090310175904.GB26351@sigill.intra.peff.net> In-Reply-To: <94a0d4530903091450gdcad625g95cd9550cfb724fa@mail.gmail.com> On Mon, Mar 09, 2009 at 11:50:09PM +0200, Felipe Contreras wrote: > 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 I don't know of a good way to check at compile time that the struct member is non-NULL. So I think we are stuck with parse_options dying, or doing something to cover it up, like: > Perhaps not showing the option at all, or perhaps showing "**EMPTY**". But the problem there is that if you _did_ mean for it to be shown, such a bug can easily escape notice. Tests will generally still pass, so unless you are scanning the output manually, nothing will tell you what happened. I think our best bet is to simply try to catch such things in code review. This one slipped through, but it was still caught in 'next', most because it _did_ crash on a non-glibc platform. > 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. Fair enough. My patch is is in 'next' now, so please go ahead and submit a new patch with your suggested change. You might want to change --bool and --int at the same time, though, as they use the same wording. -Peff