[gentoo-dev] [RFC] QA bashism check on portage

[gentoo-dev] [RFC] QA bashism check on portage

[Timothy Redaelli]

[-- Attachment #1: Type: text/plain, Size: 518 bytes --]

Hi,
what do you think about checking for bashism on install_qa_check?
Obviously only for scripts with #!/bin/sh and #!/sbin/runscript as first line.

I think checkbashisms.pl [1] could be a good start point.

[1] http://svn.debian.org/viewsvn/devscripts/trunk/scripts/checkbashisms.pl

-- 
Timothy `Drizzt` Redaelli
FreeSBIE Developer, Gentoo Developer, GUFI Staff
There are two major products that come out of Berkeley: LSD and UNIX.
We don't believe this to be a coincidence.      -- Jeremy S. Anderson

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] [RFC] QA bashism check on portage

[Mike Frysinger]

On Wednesday 25 February 2009 11:10:01 Timothy Redaelli wrote:
> what do you think about checking for bashism on install_qa_check?
> Obviously only for scripts with #!/bin/sh and #!/sbin/runscript as first
> line.
>
> I think checkbashisms.pl [1] could be a good start point.
>
> [1] http://svn.debian.org/viewsvn/devscripts/trunk/scripts/checkbashisms.pl

i recall it being incorrect in some cases (it checked for what dash supports, 
not what POSIX supports), but that was a while ago, so maybe my experience is 
dated at this point.  otherwise, integrating it sounds sane to me, and if we 
can push fixes back to debian, even better.

a repoman check on files in files/ would be useful as well i'd think
-mike

Re: [gentoo-dev] [RFC] QA bashism check on portage

[Timothy Redaelli]

[-- Attachment #1: Type: text/plain, Size: 1075 bytes --]

On Wednesday 25 February 2009 23:45:41 Mike Frysinger wrote:
<cut>
> i recall it being incorrect in some cases (it checked for what dash
> supports, not what POSIX supports), but that was a while ago, so maybe my
> experience is dated at this point.  otherwise, integrating it sounds sane
> to me, and if we can push fixes back to debian, even better.

It does not fail, the problem is that some construct works only on some shells 
or systems (for example type and kill -HUP warnings).

Maybe we should add a getopt flag (-q for example) to mask that types of 
warnings, or maybe we should (for Gentoo/Alt) adapt files to be (almost) 
POSIX compliant (so command -v instead of type, kill -s HUP instead of 
kill -HUP, and go on)

> a repoman check on files in files/ would be useful as well i'd think

Yes, as supplementary check

-- 
Timothy `Drizzt` Redaelli
FreeSBIE Developer, Gentoo Developer, GUFI Staff
There are two major products that come out of Berkeley: LSD and UNIX.
We don't believe this to be a coincidence.      -- Jeremy S. Anderson

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] [RFC] QA bashism check on portage

[Mike Frysinger]

On Thursday 26 February 2009 04:27:44 Timothy Redaelli wrote:
> On Wednesday 25 February 2009 23:45:41 Mike Frysinger wrote:
> > i recall it being incorrect in some cases (it checked for what dash
> > supports, not what POSIX supports), but that was a while ago, so maybe my
> > experience is dated at this point.  otherwise, integrating it sounds sane
> > to me, and if we can push fixes back to debian, even better.
>
> It does not fail, the problem is that some construct works only on some
> shells or systems (for example type and kill -HUP warnings).
>
> Maybe we should add a getopt flag (-q for example) to mask that types of
> warnings, or maybe we should (for Gentoo/Alt) adapt files to be (almost)
> POSIX compliant (so command -v instead of type, kill -s HUP instead of
> kill -HUP, and go on)

i'm totally not following.  we were talking about POSIX shell syntax, but now 
you're talking about utilities as well ?
-mike

Re: [gentoo-dev] [RFC] QA bashism check on portage

[Mike Frysinger]

On Thursday 26 February 2009 06:33:17 Timothy Redaelli wrote:
> On Thursday 26 February 2009 10:32:52 you wrote:
> > i'm totally not following.  we were talking about POSIX shell syntax, but
> > now you're talking about utilities as well ?
>
> I'm talking about checkbashism.pl checks and btw type is a builtin and kill
> can also be a builtin

it doesnt matter if they're builtins, they still need to follow POSIX

> kill -HUP
> `cat /var/run/mDNSResponderPosix.pid` >&/dev/null possible bashism in
> /etc/init.d/mDNSResponderPosix line 35 (kill -[0-9] or -[A-Z]): kill -HUP

POSIX states that `kill [-signal_name]` is valid syntax.  if the shell doesnt 
support it, then it isnt POSIX complaint and therefore it sucks and is not 
worth consideration.
-mike