On 17/07/2020 10:02, Gert Doering wrote:
> Acked-by: Gert Doering <g...@greenie.muc.de>
> 
> I have not tested the actual docutils / openvpn.8 generation (Samuli will 
> complain loudly if tarball making doesn't work anymore, so that *will* 
> see testing).  Generally it looks sane.
> 
> This condition looks a bit fishy, though...
> 
>  +AM_CONDITIONAL([HAVE_PYDOCUTILS], [test "${RST2MAN}" -a "${RST2HTML}"])
> 
> not sure what that will do in a POSIX shell.

Hmm ... whoops.  That should probably have been

    test -n "${RST2MAN}" -a -n "${RST2HTML}"

Not sure how that passed during my own tests.  I tested it on a various of
boxes, but I only have Linux distros easily available.

> Maybe this shouldn't be conditional either
> 
>  +if HAVE_PYDOCUTILS
>   dist_noinst_DATA += openvpn.8
> 
> because it will lead to "tarballs randomly contain openvpn.8 or not, 
> depending on whether docutils are around" - "make dist" should behave
> consistently, and if there are no docutils, I think it should fail, not
> silently leave out files.

The intention is that the tarball contains prebuilt openvpn.8 and openvpn.html
files, which is generated by "make {dist,distcheck}".  If these files exists,
they will not be rebuilt unless explicitly removed.  So most users building
from the source tarball should not notice any difference from prioer OpenVPN
releases.  This is what the additional dist-hook rule in doc/Makefile.am does;
this is run right before the copied source tree is put into a tarball.

The challenge is that it must be a conditional to actually pass ./configure -
even when built from source tarballs.  Otherwise python-docutils will be a
build dependency.  We discussed this at the Trento Hackathon and you where
skeptic to require a Python stack to be installed to build OpenVPN (as that
stack is not a common system dependency in the non-Linux world).  So we agreed
we will ship pre-built man/html files in the source tarballs.

However, to do the "make {dist,distcheck}" from the *git repo*,
python-docutils need to be a mandatory dependency - because we don't check in
the prebuilt openvpn.8 and openvpn.html files into the git repo.

This logic could probably contains some flaws and can be further improved, but
I figured we need to get this tested on a broader set of use cases and
OS/distros to better see which annoyances which hits us.


-- 
kind regards,

David Sommerseth
OpenVPN Inc


Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
Openvpn-devel mailing list
Openvpn-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/openvpn-devel

Reply via email to