Re: [PATCH] Revert "dpkg: sbuild allows extra arguments via DPKG_SBUILD_EXTRA_ARGS"

[Henning Schild <henning.schild@siemens.com>]

Am Thu, 2 Feb 2023 23:44:46 +0100
schrieb "Roberto A. Foglietta" <roberto.foglietta@gmail.com>:

> On Thu, 2 Feb 2023 at 21:14, Henning Schild
> <henning.schild@siemens.com> wrote:
> >
> > Am Thu, 2 Feb 2023 20:06:12 +0100
> > schrieb "Roberto A. Foglietta" <roberto.foglietta@gmail.com>:
> >  
> > > On Thu, 2 Feb 2023 at 15:42, Jan Kiszka <jan.kiszka@siemens.com>
> > > wrote:  
> > > >
> > > > From: Jan Kiszka <jan.kiszka@siemens.com>
> > > >
> > > > This reverts commit 6d5fbbab9c9e3c894cbc9cf18180fb0057f82f20.
> > > >
> > > > There are no concrete use cases for this interface known.
> > > > Adding it without an idea what it should be good for can create
> > > > downstream problems as the class evolves and also motivate
> > > > inappropriate use of this low-level interface in recipes.  
> > >
> > > Open-source is about empowering the users and unless illegal use
> > > of it, there is no any reason to judge the use they do of it. It
> > > is hard to think that a variable would motivate an "inappropriate
> > > use" unless the user did not find a way to address a problem and
> > > the go to look in that class. At that point, it might find a
> > > quick solution in that variable without the need of overload the
> > > entire class in the top layer and modify the class loosing the
> > > upstream updates. I agree that there is no reason to documenting
> > > that variable. It is self-explanatory and quick to identify for
> > > those get deep into the dpkg class.  
> >
> > All EXTRA_ARGS switches are very dangerous because they allow very
> > unintended hacks. We have several of them and some of them have
> > caused some serious fun. Especially if people use them
> > "code-injection-style"
> >
> > like EXTRA = "; true; sudo fun"
> >  
> 
> Some people might play funny tricks in their projects and no one could
> prevent this but what has to do with ISAR? Why do we need to worry
> about that? Overloading the class or function would not allow them to
> play any trick they want? They can append or prepend whatever they
> want in every class and every function they want. Why are extra
> arguments variables such a problem and :append or :prepend or
> .bbclassappend are not?

Yes bitbake allows all tricks, so why merge your patch?

It has to do with Isar because you suggest a new variable/interface
that Isar will have to maintain. If you refuse to put the problem on
the table ... there is no reason to take your patch. And no
abuse-potential and no need to maintain.

Henning

> Best regards, R-