Re: Bug#724469: FTBFS on all big-endian architectures

[intrigeri <intrigeri@debian.org>]

Hi Perl team,

intrigeri wrote (26 Mar 2014 11:25:38 GMT) :
>  2. I want Jessie to be released with this package (and its
>     reverse-dependencies) working on as many supported architectures
>     as possible. Given I don't have the skills needed to port it to
>     big-endian 64-bit architectures myself, all I can do is #1. So,
>     whether Jessie ships this package on these architectures does not
>     depend much on me.

... and I would like to have opinions about what is an appropriate
timing for dropping a target architecture for a given package, when
nobody comes up with a patch to port the code to that architecture.

In this case:

  2013-09-24: test failure on big-endian architectures reported in Debian
  2013-10-16: test failure on big-endian architectures reported upstream
  2013-11-21: initial patch for 32-bit BE architectures proposed
  2014-01-17: upstream considers the initial patch not good enough
  2014-01-20: I invite the Debian porters to follow-up on the upstream BTS
  2014-01-22: upstream proposes another patch
  2014-01-24: a Debian porter reports the patch works on 32-bit BE
  2014-01-29: a Debian porter reports the patch does not work on 64-bit BE
  2014-02-09: upstream proposes another patch
  2014-02-21: I ping porters on the upstream BTS
  2014-03-22: given the lack of testing reports, I ask Debian porters
              to test latest upstream patch on all BE architectures,
              and get plenty of replies this time
  2014-03-23: I apply the patch that fixes things on 32-bit BE,
              upload, and ask upstream and the 64-bit BE porters if
              they plan to work on the 64-bit BE side (my wording was
              not appropriate, given the replies I got)

I'm a bit scared by the last exchanges I had with porters on this bug,
so I'm willing to handle this matter as gently as possible.

E.g. I'm inclined to think that people were waiting for a resolution
for 32-bit BE to be tested and accepted, before starting to work on
the 64-bit side. So, probably the counters are now somewhat reset.

At the same time, I won't wait forever. If nobody comes up with
a patch for 64-bit big endian architectures within N months from now,
then I'll probably want to drop s390x from the list of target
architectures for this package. So, my (currently 100% theoretical)
question is: what would be an appropriate value for N?

Cheers,
-- 
  intrigeri
  | GnuPG key @ https://gaffer.ptitcanardnoir.org/intrigeri/intrigeri.asc
  | OTR fingerprint @ https://gaffer.ptitcanardnoir.org/intrigeri/otr.asc