On 21 April 2016 at 03:53, Jakub Wilk <jw...@debian.org> wrote:
> Policy §10.4 says that /bin/sh script need to follow SUSv3 + some
> extensions. Both "hash" and "command -v" are optional in SUSv3 (but they are
> mandatory in SUSv4), so in theory you shouldn't use them. In practice,
> however, "command -v" is probably more portable than anything else.
>
> "which" is not standardized, but you can rely on it as far as Debian in
> concerned, because it's shipped by an essential package.
>
> Related bugs: #747320, #733511
Thank you very much for the reference. Also, this is the first time
I've seen the section symbol "§". So just to clarify, both "command
-v" and "which" are acceptable for Debian work, and which one I use is
a matter of preference? I consulted the Policy, which recommended
checking the script with "checkbashisms." Checkbashisms reports:
possible bashism in btrfs-progs.postinst line 7 ('command' with option
other than -p)
I assume this can be disregarded, but it makes me wonder if "which" is
most Debian-correct in the strictest sense, because it can be relied
on "because it's shipped by an essential package" and because there is
no possible conflict with Policy § 10.4.
>> In this case, I think it broke on piuparts because piuparts used bash; so
>> I think this was an dashism rather than a bashism ;-)
>
>
> You can't rely on the output format of argumentless "hash"; but other than
> that it works fine in both dash and bash.
Guilty as charged!
Cheers,
Nicholas
