On 2019/07/15 09:47, Kurt Mosiejczuk wrote:
> Based on a conversation with afresh1 about the pypi DESCR, this patch
> has portgen write the description to DESCR.new rather than just
> DESCR. This means if one doesn't intervene and deal with DESCR.new,
> portcheck will complain. This is related to my forgetting that I hadn't
> checked DESCR on a new port but thought I had. Perhaps the portgen
> module in question generated a perfect DESCR, but one should make sure
> it did so.
I'm not a fan of this, it adds an extra "mv" step that needs to
be done before installing the software to test and I don't think it will
actually help anything. So the process which is currently
work on port
test/revise/retest
forget to update DESCR
run portcheck, no problem noticed
send port out
will become
work on port
try to test
can't package, mv pkg/DESCR.new pkg/DESCR, "i'll update it when i know
what the software actually does"
test/revise/retest
forget to update DESCR
run portcheck, no problem noticed
send port out
If you want portcheck to trip on it, how about writing an identifiable
string to DESCR ("* Generated by portgen: needs review" or similar) and
having portcheck look for that instead?
(People should try to remember to read every file in the port for a
quick review before they send it to ports@. When I remember - most but
not all of the time! - I just open the tar itself in vim and work through
the files in order).
> Now, this leaves the mechanism that renames the original DESCR to
> DESCR.orig for an updated port (and also does so for Makefile). I figure
> in the case of an update, doing this will also encourage looking at the
> old DESCR versus the new version. I'd be interested to know how others
> feel about that though.
As long as it doesn't change DESCR but just adds a new file this sounds
reasonable.
