Re: [PATCH] Use full path to dnsdomainname and domainname in scripts/mkcompile_h

[Glenn Sommer]

2010/1/27 Michal Marek <mmarek@xxxxxxx>:
> On Wed, Jan 27, 2010 at 10:44:29AM +0800, AmÃrico Wang wrote:
>> On Wed, Jan 27, 2010 at 3:10 AM, Glenn Sommer <glemsom@xxxxxxxxx> wrote:
>> > --- scripts/mkcompile_h.orig    2010-01-26 18:59:37.000000000 +0100
>> > +++ scripts/mkcompile_h 2010-01-26 20:03:42.000000000 +0100
>> > @@ -67,9 +67,9 @@
>> >   echo \#define LINUX_COMPILE_BY \"`whoami`\"
>> >   echo \#define LINUX_COMPILE_HOST \"`hostname | $UTS_TRUNCATE`\"
>> >
>> > -  if [ -x /bin/dnsdomainname ]; then
>> > +  if [ `command -v dnsdomainname 2> /dev/null` ]; then
>> >     domain=`dnsdomainname 2> /dev/null`
>> > -  elif [ -x /bin/domainname ]; then
>> > +  elif [ `command -v domainname 2> /dev/null` ]; then
>> >     domain=`domainname 2> /dev/null`
>> >   fi
>> >
>>
>> No, this doesn't look good.
>>
>> First, you don't need to redirect stderr for 'command'.
>>
>> Second, 'command' also searches in shell built-in commands, aliases,
>> so I prefer 'whereis -b'.
>
>
> Well, 'command -v domainname' returns success iff 'domainname' can be
> executed (be it an external command, builtin, function, whatever), which
> is exactly what we do on the next line. But, there is no need to capture
> the output of 'command -v domainname' and pass it to [ ... ], just test
> the return code.

> ... crap, now I learned that busybox doesn't support 'command' :-(
> So what about simply trying 'dnsdomainname' and falling back to
> domainname if it fails? Like this:


Ohh, I didn't know that! We DO need to be compatible with busybox! :/


>
> Subject: [PATCH] scripts/mkcompile_h: don't test for hardcoded paths
>
> Don't test for /bin/{dnsdomainname,domainname}, simply try to execute
> the command and check if it returned something.
>
> Reported-by: Glenn Sommer <glemsom@xxxxxxxxx>
> Signed-off-by: Michal Marek <mmarek@xxxxxxx>
> ---
>  scripts/mkcompile_h |    5 ++---
>  1 files changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/scripts/mkcompile_h b/scripts/mkcompile_h
> index 23dbad8..50ad317 100755
> --- a/scripts/mkcompile_h
> +++ b/scripts/mkcompile_h
> @@ -67,9 +67,8 @@ UTS_TRUNCATE="cut -b -$UTS_LEN"
>   echo \#define LINUX_COMPILE_BY \"`whoami`\"
>   echo \#define LINUX_COMPILE_HOST \"`hostname | $UTS_TRUNCATE`\"
>
> -  if [ -x /bin/dnsdomainname ]; then
> -    domain=`dnsdomainname 2> /dev/null`
> -  elif [ -x /bin/domainname ]; then
> +  domain=`dnsdomainname 2> /dev/null`
> +  if [ -z "$domain" ]; then
>     domain=`domainname 2> /dev/null`
>   fi
>
> --
> 1.6.5.3
>
>

I tested above patch, and it seems to work fine.
Though, by looking a bit closer at the source - I found we actually
NEVER need to use 2> /dev/null.
We capture the output using ") > .tmpcompile", meaning we only capture
stdout _NOT_ stderr.
One could argue that we should remove the redirection of stderr for
debugging purposes. (In really rare cases where both dnsdomianname and
domainname are missing)
Though, I do NOT see the need for that!

I'm completely satisfied with the above patch! It does the job I
initially requested - which was to either use a complete path at all
times, or never use a complete path. :)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/