Re: [PATCH 8/8] diff-highlight: fetch all config with one process
- From
Jeff King <peff@peff.net>
- Date
- Mar 22, 2026, 20:45 UTC
- Message-ID
- <20260322204509.GA2047044@coredump.intra.peff.net>
- In-Reply-To
- <9d3633e4-6413-4932-a29d-e0347546ede8@malon.dev>
On Mon, Mar 23, 2026 at 01:18:30AM +0800, Tian Yuchen wrote:
Show 10 quoted lines
> > - my $s = `git config --get-color $key 2>$NULL`;
> > +
> > + if (!defined $cached_config) {
> > + $cached_config = {};
> > + my $data = `git config --type=color --get-regexp '^color\.diff-highlight\.' 2>$NULL`;
> > + for my $line (split /\n/, $data) {
> > + my ($key, $color) = split ' ', $line, 2;
>
> ...another $key. I think it would be better to change the name here. What do
> you think?I noticed it, too, but didn't have a better name (in fact they are of the same type, just two different contexts). Shadowing seemed less bad to me than using a mis-matched name.
Show 10 quoted lines
> > + my $s = $cached_config->{$key};
> > return length($s) ? $s : $default;
> > }
>
> Something doesn't feel quite right here.
>
> If the user has not configured color.diff-highlight.*, the expression git
> config --type=color --get-regexp=^color\.diff-highlight\. will not find a
> match and should not output anything. In this case, %cached_config->{$key}
> becomes undef, length() returns 0, and a warning is issued.The length() of undef is also undef (and documented in "perldoc -f length"). But either way, length($s) will be false, and we will return $default, not $s.
I don't get any warning on perl 5.40.1. Are you seeing one on a different version?
-Peff