Re: [Debian-med-packaging] r12212 - trunk/packages/libsbml/trunk/debian

[Steffen Möller <steffen_moeller@gmx.de>]

On 09/15/2012 11:13 AM, Andreas Tille wrote:
> Hi Steffen,
> 
> On Fri, Sep 14, 2012 at 08:26:22PM +0000, Steffen Möller wrote:
>>
>> Modified: trunk/packages/libsbml/trunk/debian/rules
>> ===================================================================
>> --- trunk/packages/libsbml/trunk/debian/rules	2012-09-14 13:57:12 UTC (rev 12211)
>> +++ trunk/packages/libsbml/trunk/debian/rules	2012-09-14 20:26:22 UTC (rev 12212)
>> @@ -105,7 +105,7 @@
>>  	### verx ugly ###
>>  	rm -rf .pc .debs config docs examples macosx src config.guess config.sub
>>  	dh_clean
>> -	tar xzf ../libsbml_$(version).orig.tar.gz -C ../
>> +	tar --no-same-owner -xzf ../libsbml_$(version).orig.tar.gz -C ../
> 
> I guess you consider that --no-same-owner is fixing my remark about
> changing md5 sum of tarball

I was not aware of any email of yours on the package. A tar executed
as root (and apparently also as fakeroot) attempts to preserve user IDs.
The "--no-same-owner" is the default setting for regular users.

> but it remains totally wrong

I did not like the tar in there either, indeed. It should be fixed,
I agree. It was not dramatic, though, and more of an uptream issue
that we could indeed think about addressing at some point. The
clean target can expect the tarball to be existing, tested also
with pbuilder, and if not violating the policy, I did not see why
a maintainer should not use what it available.

> and you should
> definitely not have uploaded the package.  You should never ever create
> the orig.tar.gz in a clean target and I wished that this knowledge is
> spread between all team members instead of uploading inacceptable dirty
> hacks.

Aaaah, you apparently misread the "x" for a "c" .  That would have
crossed a red line, indeed. The untar was still sufficiently gr[ea]yish
for me.

Steffen