From: Patrick Steinhardt <ps@pks.im>
Date: Fri, 09 Jan 2026 16:32:39 GMT
Subject: Re: [PATCH v17 1/2] refactor format_branch_comparison in preparation
Message-ID: <aWEtp77UB30SbKle@pks.im>
In-Reply-To: <848006C9-FD1B-4E73-935F-D4A338AF2EC9@gmail.com>

On Fri, Jan 09, 2026 at 11:22:33AM -0500, Ben Knoble wrote:
> 
> > Le 9 janv. 2026 à 11:07, Harald Nordgren <haraldnordgren@gmail.com> a écrit :
> > 
> > 
> >> 
> >> Using an enum for a set of flags is a bit confusing.
> > 
> > The point of the flag and the bitmasking is to selectively turn off the push
> > and pull advice advice from the relevant branch when the push branch
> > comparison is active.
> > 
> > In an earlier implementation the advice logic was moved to the caller
> > instead 'format_branch_comparison', but it's more faithful to the original
> > to have the advice logic inside 'format_branch_comparison'. Maybe I
> > misunderstood your comments around this?
> > 
> > Would happily take a suggestion on a nicer way to handle it.
> 
> I think other uses in Git declare a bunch of integer constants for the
> bitfields, not an enum. Why? Because or-ing flags together creates a
> value not in the enum…

We nowadays typically declare the flags as enum, but when accepting the
bitfield we use `unsigned`.

Patrick