Simon Josefsson wrote:
> > There are actually 3 ways to use gnulib:
> > 1) manual file copies, considering the contents of the module
> > descriptions,
> > 2) using a script based on the elementary --extract-... accessors of
> > gnulib-tool,
> > 3) using gnulib-tool --import.
>
> Do we want to guarantee that gnulib should be usable in all these
> ways?
Jim has expressed the opinion that "yes". I'm currently convinced as well
that it is useful to keep the transparency that makes 1) possible: People
trust something more if they can understand the bricks and if there's not
too much magic in it.
> In other words, would it be useful to set up an autobuilder
> that automates testing of the 3 approaches
How could this be possible?
1) How do you make an autobuilder that "manually considers the contents
of the module descriptions"?
2) Here you would need to write a different script that performs the
same effect as "gnulib-tool --import". I consider this a waste of time.
In summary, from the 'users' file we see that the vast majority is using
gnulib-tool, and therefore IMO this is the only approach worth testing.
Bruno
