On 2021-06-29 20:36:36-0700, Junio C Hamano <gitster@xxxxxxxxx> wrote:
> Eric Sunshine <sunshine@xxxxxxxxxxxxxx> writes:
>
> > I may be misunderstanding your suggestion, but isn't the proposed
> > test_output_wc_l() function the same as what Danh had originally
> > implemented several re-rolls back (though he named it
> > test_line_count_cmd())?
>
> Could be, except that I recall we saw extra noises like --out/--err
> that weren't used and contaminating the current working directory,
> which are gone from the latest iteration.
Yes, in v{1,2}, there's the extra noises of --out and --err,
but it's gone in v3.
I guess you're thinking about the contamination of $PWD iff it's not
a git worktree. That could be simplified by BUG-ging if $PWD is not
a git worktree.
Maybe, you're thinking about the extra noises to handle the failure run
of command under check? That could be dropped, too.
Would you mind looking at v3 1/4 again to see what is your opinion
there? I don't mind re-roll a same or simplified version of v3,
with s/test_line_count_cmd/test_output_wc_l/ if you see fit.
> The simplification
> compared to that iteration is quite welcome---it made the resulting
> code that uses the helper easier to follow compared to the earlier
> attempts. But this round simplifies it too much and the results got
> harder to follow by burying the command name in the helper and made
> it less obvious that the last part of the helper's parameters are
> arguments given to ls-files, I would think.
--
Danh
