[Buildroot] [RFC PATCH 0/2] use `command -v' instead of `which'

[Petr Vorel]

Hi Yann, all,

> Petr, Arnout, All,

> On 2021-09-26 23:32 +0200, Arnout Vandecappelle spake thusly:
> > On 21/09/2021 22:51, Petr Vorel wrote:
> > >I've tested the patchset on dash as the default shell. But it certainly
> > >deserve more people to have look and test.

> >  Well, as the commit message says: it's POSIX so it should be supported by
> > everything. which has a much smaller chance of being supported.

> This is causing quite some issues.

> First, 'command -v' does not behave the same way 'which' used to, when
> passed more than one parameter, because some shells are not compliant to
> POSIX (this might be a bug, but nonetheless it affects the most widely
> used shell out there, bash) [0].
Yes, there has been a proposal, how to fix this.

> Second, this is causing a lot of error messages:

>     $ make defconfig
>     [...]
>     $ make help
>     make[1]: command: Command not found
>     [...]
New error. But I was not able to reproduce it on x86_64 on current master
(5916cc5011). What am I missing to reproduce it?

> The original commit reports that 'which' is broken in Debian, but I was
> not able to reproduce in Bullseye, where 'which' still works as expected
> and does not emit any extra warning.
Yes, the waring is in debianutils 5.x, which is still in Debian unstable (not
even in the testing) [1].

> So, we are trying to fix something that is broken on a development
> version of Debian, but that still works in all known released
> distributions.
'command -v' and 'type' has been here quite long time as well.

> I usually am quite in favour of sticking to POSIX tools, but which has
> been ubiquitous in the past 30 years or so, and I would consider that
> Debian's which *is* broken for reporting such deprecation warnings.

> So, I suggest that we do revert this patch, and work on a better
> transition away from which, if at all. One very quick solution would be
> to bundle our own which in Buildroot and then we'd have a quick way out
> of that Debian's mess...
Sure, if it causes problems which are not easily fixed, I'm not against
reverting it. But I don't think that problem is that complex, that we'd need to
compile which. But I apologize for causing troubles.

Kind regards,
Petr

> Anyway, I'd vote "revert".

> Regards,
> Yann E. MORIN.

> [0] https://lore.kernel.org/buildroot/YVTIghzHs82uFBIe@pevik/T/#m95c17eb8374e4e3dd6eee700d397aa12cca0739e
[1] https://tracker.debian.org/pkg/debianutils