Re: [PATCH v2 1/6] sha1-file.c: remove the_repo from read_object_with_reference()
- From
Johannes Schindelin <johannes.schindelin@gmx.de>
- Date
- Jun 28, 2019, 12:46 UTC
- Message-ID
- <nycvar.QRO.7.76.6.1906281440400.44@tvgsbejvaqbjf.bet>
- In-Reply-To
- <20190627092852.11326-2-pclouds@gmail.com>
Hi Duy,
On Thu, 27 Jun 2019, Nguyễn Thái Ngọc Duy wrote:
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> > ---
This commit is _awfully_ short given that...
Show 11 quoted lines
> diff --git a/builtin/grep.c b/builtin/grep.c > index 580fd38f41..560051784e 100644 > --- a/builtin/grep.c > +++ b/builtin/grep.c > @@ -458,7 +458,8 @@ static int grep_submodule(struct grep_opt *opt, > object = parse_object_or_die(oid, oid_to_hex(oid)); > > grep_read_lock(); > - data = read_object_with_reference(&object->oid, tree_type, > + data = read_object_with_reference(&subrepo, > + &object->oid, tree_type,
... this change and...
Show 10 quoted lines
> &size, NULL); > grep_read_unlock(); > > @@ -623,7 +624,8 @@ static int grep_object(struct grep_opt *opt, const struct pathspec *pathspec, > int hit, len; > > grep_read_lock(); > - data = read_object_with_reference(&obj->oid, tree_type, > + data = read_object_with_reference(opt->repo, > + &obj->oid, tree_type,
... this change is totally not what would be intuitively the easiest: to use `the_repository` in all built-ins.
It might take quite a lot of convincing that these changes are correct, in particular in light of the regressions introduced by the first iteration (to paraphrase Warren Buffet [*1*]: one slip in a patch series touching as central parts as this one will need a lot of time to restore trust in subsequent iterations' correctness.)
In short: with such an empty commit message, this patch is no good. It's as if it was optimized to pass the test suite on Linux instead of a best effort to make the conversion as correct as you can make it.
Ciao, Johannes
Footnote *1*: https://www.forbes.com/sites/jamesberman/2014/04/20/the-three-essential-warren-buffett-quotes-to-live-by/