> It's not a good idea to have the `make refdoc` target call pip to
> install missing requirements in the user's python lib.
Hmm. `make refdoc' is not a target that is to be called by Joe User.
It was always handled specially, i.e., it was assumed, for example,
that Python is available, without testing for its existence.
> I don't want those packages to pollute my environment, if I just
> want to build the freetype documentation.
>
> You should either exit with error if the requirements are not met,
> and maybe tell the user to run pip install -r requirements.txt for
> themselves;
Yes. Nikhil, please replace the call to `pip install' with a check
whether the necessary packages are available, emitting a list of
packages that the user has to install to proceed.
> or you could bootstrap a new python virtual environment, [...]
I don't have an opinion here since I'm not Python user. My gut
feeling, however, says that we should first try to make it run with
installed packages also.
> This would also avoid issues like the one Werner reported where
> there is a conflict between the version of Pygments installed
> globally (via the distro package manager) and the one required by
> the docwriter.
Hmm, I don't see a version test for Pygments in `requirements.txt'.
Nikhil?
Werner
_______________________________________________
Freetype-devel mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/freetype-devel
