On Fri, Nov 20, 2015 at 06:05:00PM +0000, Rainer Weikusat wrote:
> Timo Buhrmester <fstd.lkml@???> writes:
> >> >> >> > Fatal error: can't create obj/backend.o: No such file or directory
> >> >> >> create a dummy file in this directory and check that into the SCM.
> >> >> > Why should the repository have to remember a build system detail?
> >> >>
> >> >> Because the build system is part of the code.
> >> > It would still be preferable if the build system was stand-alone, rather
> >> > than depending on a repository twist.
> >>
> >> It would be "preferable" if a SCM would just store whatever one desires
> >> to put into it.
> > Do you have any point or do you just enjoy typing?
>
> One could argue that I was somewhat careless in my first reply to you as
> I actually wrote something about you (namely, "that's your opinion but
> not mine") which sort-of solicitated an angry retort.
???

> However, stating that "a source code management system" or - better term
> here - "version control system" should simply store whatever someone
> desires to store in, just like - say - a filesystem, is certainly not
> a personal remark.
Correct, that is not a personal remark. Why are you pointing this out?


Just to get back on topic since you obviously lost track of the discussion:
- We have a build system. Let's call it build.sh for simplicity, regarless of
what it is in reality.

- build.sh is set to build the project in a directory called "obj".

- "obj" has no other purpose than to hold the results of the build process.

Who should create "obj", and why?

I feel that using a repository hack to keep obj around introduces
an unnecessary and trivially avoidable dependency. What if I don't
obtain the source through a check-out of that repository?
What is the benefit of /not/ having build.sh mkdir obj?