[BusyBox] [PATCH] New comm applet

[Rob Landley]

On Wednesday 05 April 2006 4:09 am, Bernhard Fischer wrote:
> On Tue, Apr 04, 2006 at 02:28:33PM -0500, Rob Landley wrote:
> >On Monday 03 April 2006 5:54 pm, Bernhard Fischer wrote:
> >> On Sat, Apr 02, 2005 at 05:55:17AM -0500, Rob Sullivan wrote:
> >> >Hi,
> >> >from reading this mailing list, I discovered that comm was needed to
> >> >build Perl, and that no version was currently available in Busybox.
> >>
> >> Could someone who is using comm look whether these are ok or not?
> >>
> >> coreutils/comm.c:78: warning: the address of 'thisline', will always
> >> evaluate as 'true'
> >> coreutils/comm.c:85: warning: the address of 'thisline', will always
> >> evaluate as 'true'
> >
> >I think they meant to test the value at that location.  (That's what I
> > changed it to do, anyway.)
>
> If this is something you (Rob S.) want to have in 1.1.x, then please
> test and ACK. It was r14748:
> http://busybox.net/downloads/patches/svn-14748.patch
>
> TIA.

Nobody's actually reported being inconvenienced by comm not working for them, 
and I'd prefer somebody who has an actual use case for comm to test it before 
twiddling stable.  My fix made the warnings go away, but they weren't really 
based on the logic of what the program was doing, just what would be a common 
mistake to get that warning.  For all I know, the changes broke it rather 
than fixed it.

The only real world test I know of is the perl build, and I haven't built perl 
in a while.  (I don't use it, and my Firmware system doesn't need it.)  
Speaking of which, I should probably move firmware linux development to 
busybox.net/~landley, since it's my test case for busybox the way buildroot 
is Erik's testcase for uClibc.  (I'd set up an svn for it, but who else is 
interested?)  On an unrelated note, I got an arm kernel built from source to 
boot under qemu's arm system emulation yesterday.  Life is good.

Anyway, getting back to comm: if somebody wanted to put together a comm.tests 
from susv3, that would be good enough for me to put the patch into -stable...

http://www.opengroup.org/onlinepubs/009695399/utilities/comm.html

I believe "uniq" or "sum" would be good examples of complete small test cases.  
(Blatantly, _shamelessly_ asking other people to do work.  Not expecting it 
to happen, I'll get around to it eventually myself.  But not this week.)

Rob
-- 
Never bet against the cheap plastic solution.