From: Pierre Habouzit Date: Wed, 28 Jul 2010 13:06:11 GMT Subject: Re: [RFC PATCH 0/2] Allow detached forms (--option arg) for git log options. Message-ID: <20100728130610.GG6895@madism.org> In-Reply-To: On Tue, Jul 27, 2010 at 08:10:35AM -0700, Jakub Narebski wrote: > Pierre Habouzit writes: > > > On Mon, Jul 26, 2010 at 02:31:09PM -0500, Jonathan Nieder wrote: > > > Hi Matthieu, > > > > > > Matthieu Moy wrote: > > > > > > > is there any reason why "git log" hasn't been > > > > migrated to parse-option? Or is it only that nobody did it yet? > > > > > > Please go ahead. :) > > > > I started it in the past, but never went around to actually do it. > > > > I started to get rid of most of the bitfields to use explicit or-ed > > fields, but stopped at that, I don't even remember if those patches got > > merged or not. > > Why did you feel this change was needed / necessary? Was it > limitation of parseopt? Or perhaps it was for portability reasons? > Or was it just the matter of code elegance? you cannot take the address of a bit portably in C, so you can't let parseopt set/clear bits through bitfields (as in unsigned field : 1 in a struct in C I mean). So to use parseopt OPTION_BIT feature, you have to convert them to C flags as in "unsigned flags" and explicit masks defines/enums. IOW: struct foo { unsigned bar : 1, ... baz : 1; }; Must be converted into: struct foo { #define FOO_FLAG_BAR (1U << 1) ... #define FOO_FLAG_BAZ (1U << 18) unsigned flags; } so that you can use parseopt. that's what I meant. This was done for the rev-list parsing stuff e.g. -- ·O· Pierre Habouzit ··O madcoder@debian.org OOO http://www.madism.org