Re: debian/rules clean as root or non-root?

[Goswin von Brederlow <goswin-v-b@web.de>]

Sven Joachim <svenjoac@gmx.de> writes:

> On 2010-02-10 21:37 +0100, Goswin von Brederlow wrote:
>
>> Sven Joachim <svenjoac@gmx.de> writes:
>>
>>> On 2010-02-10 19:02 +0100, Goswin von Brederlow wrote:
>>>
>>>> I often see sources where debian/rules clean aborts claiming it needs to
>>>> be run as root. So then I run it with fakeroot. But if the source was
>>>> previously build as root then running fakeroot debian/rules clean might
>>>> not be enough. I think the existing dh_testroot helper is insufficient
>>>> and anoying for the job.
>>>
>>> It is both insufficient and unnecessary, I don't use it in my packages.
>>> Neither does "dh clean", BTW.
>>
>> If you build the package as real root and it creates a new directory
>> (like installing into debian/package/ always does) then you can not
>> clean without being real root. So the check isn't unnecessary.
>
> Yes, and dh_clean will almost certainly fail in such a situation.
> Unless, maybe, you went to the insanity of running the build target with
> real root rights.

That is usualy what happens. Haven't yet completly tought my coworkers
to always build packages as user.

The extended dh_testroot I suggested would also catch cases where the
package was build by another user I think. Might need some tuning (chmod
770?) to allow building by multiple users in a common group and setgid
bit set.

>> If you did run only debian/rules build as root the error might
>> get ignored and lost in the output.
>
> Why would anyone want to run the build target as root?  If you do that,
> even running the clean target as root might not give you a state where
> you can build the package as a normal user again.
>
> Sven

It happens. The important part would be to clearly catch such a case. I
can then delete the working dir and unpack the source again or check it
out again.

MfG
        Goswin