Re: [PATCH v2] futex_waitv.2: new page

From: Jeff King <peff@peff.net>
Date: Feb 11, 2026, 07:35 UTC
Message-ID: <20260211073553.GA1867915@coredump.intra.peff.net>
In-Reply-To: <55bd6ad3-cd19-4839-a808-eceae802d3c8@app.fastmail.com>

On Tue, Feb 10, 2026 at 04:54:02PM +0100, Kristoffer Haugsbakk wrote:

> An aside but `format.noprefix` is not a boolean like `diff.noprefix`.
> It will be enabled with any value.

Huh, that's weird. Indeed, the code is:

          if (!strcmp(var, "format.noprefix")) {
                  format_no_prefix = 1;
                  return 0;
          }

which seems just wrong. I cannot think of any other case where the _existence_ of a config variable determines the outcome, rather than the assigned value. So I would be inclined to call this a bug and switch it to use git_config_bool().

Technically that is not backwards-compatible, but I really think the existing behavior is just a bug. It was not something intended and is contrary to how the rest of Git works. Presumably nobody noticed because why in the world would you set it to false in the first place?

Show 11 quoted lines

> Is it standard to indicate this with the existing “If set,”, perhaps? Or
> should it say “enabled with any value”?
> 
> (+Cc Peff)
> 
>     format.noprefix::
>             If set, do not show any source or destination prefix in patches.
>             This is equivalent to the `diff.noprefix` option used by `git
>             diff` (but which is not respected by `format-patch`). Note that
>             by setting this, the receiver of any patches you generate will
>             have to apply them using the `-p0` option.

We usually use "if set" in the config documentation to refer to options being set to the true value. So I think the documentation text is OK and would match the code, once fixed.

Kristoffer, do you want to produce a patch to fix the code? I feel like finding it was 99% of the work. ;)

-Peff