Re: [PATCH v2 01/11] builtin/init: rename "--ref-format=" to "--ref-storage-format="
- From
Kaartic Sivaraam <kaartic.sivaraam@gmail.com>
- Date
- Sep 9, 2026, 09:14 UTC
- Message-ID
- <472d9d2b-efea-48a9-98f0-02364b1a9ccf@gmail.com>
- In-Reply-To
- <aqED9Xn6xibcp9pf@pks.im>
On 9/9/26 12:30, Patrick Steinhardt wrote:
Show 8 quoted lines
> On Tue, Sep 08, 2026 at 02:31:22PM +0530, Kaartic Sivaraam wrote: > > I'm aiming for a a quiet deprecation -- the old options keep on working, > but we don't show them anymore. We may eventually remove them > altogether, but I neither want to warn about use of old options now or > even remove them as that would likely cause more negative consequences > than is worth it. It's not like we want to soon use the old names. >
Understood.
Show 5 quoted lines
> > But I agree that we should keep the old options documented. Otherwise, > it might be hard to figure out what those options do in case they for > example still exist in an old script. >
Yeah, makes sense.
Show 9 quoted lines
>> >> Relevant only if we are deprecating the old option: We could change the >> description of the old option to convey that it is deprecated so that it >> could be easily distinguished by anyone reading through. I suggest this by >> noting a similar pattern in `builtin/name-rev.c` for the `--stdin` argument. > > We don't really have to update the description as we pass > `PARSE_OPT_HIDDEN` anyway, so the option is not even shown. >
Indeed. I was suggesting to change the description only for the sake of someone who reads the code to quickly get an idea that we have deprecated the old names. The PARSE_OPT_HIDDEN would definitely hint at it but this just a suggestion to make it more obvious :-)
That said, it's fine if we don't change the description too. It isn't a big deal.
Show 9 quoted lines
>> If we don't plan to deprecate, I suppose we could use OPT_ALIAS to clarify >> that `ref-format` is an alias of `ref-storage-format` similar to how >> `recursive` is marked as an alias of `recurse-submodules` in >> `builtin/clone.c`. > > I was originally planning to use `OPT_ALIAS()`, but we don't seem to > support `PARSE_OPT_HIDDEN` there. Maybe it's better to add a small > preparatory patch to support that though. >
Indeed. OPT_ALIAS does have that caveat. A patch to allow the alias to be hidden by choice would indeed be worthwhile.
-- Sivaraam