On 27.06.2016 [12:11:41 -0000], Robie Basak wrote:
> Careful. If the landing of one SRU breaks another package, it isn't
> sufficient just to SRU both. A Breaks: needs to be added so that users
> don't accidentally pick up one SRU without the other. See bug 1511735
> for an example of how this can go wrong.
Thank you for noticing this, Robie, very good points.
It's not a real "Breaks" in this case, thankfully.
> If I understand this correctly, what you want to do is:
>
> SRU php-horde-mapi
> SRU this phpseclib with a Breaks: php-horde-map (<< version-just-SRUd)
>
> Then ask the SRU team to land both together. Though with the Breaks, apt
> will generally do the right thing if both don't land together, although
> it still could confuse users ("why won't phpseclib update?").
>
> I assumed that php-horde-mapi would actually be broken at runtime
> though, as opposed to a test positive only. If it's not broken at
> runtime, then I guess the consequence isn't so severe. I tend to fall on
> the side of fixing the dep8 test in an SRU anyway though, as otherwise
> the test becomes useless in detecting SRU regressions.
I'ts just a test positive issue, because PHP7 emits a warning to stderr
about deprecation (in src:phpseclib) of same-named constructors during
the test of php-horde-mapi. We can't fix phpseclib, as that causes
regressions like this one, so Debian (and 16.04) have add
src:php-phpseclib, which is actually v2 of phpseclib and is PHP7
compliant. Most packages have moved forward to v2, but not all, in
Stretch, so the older package still exists.
So, if I understand the process correct: this phpseclib update can be
uploaded. It will get stuck in -proposed because php-horde-mapi will
regress its tests. Once we have a proper fix in php-horde-mapi, we can
SRU that back and both will go through. However, given the SRU
timelines, etc., it might be prudent to wait for the php-horde-mapi fix
to be available before SRU'ing either.
Note also there is a much smaller fix that we have pushed to
php-horde-mapi in 16.10, so that phpseclib could sync in 16.10, and that
is to just add a
Restrictions: allow-stderr
to debian/tests/control, so the deprecation warning is not treated as a
failure.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1574058
Title:
php-seclib: Call to undefined method Crypt_Base::Crypt_Base()
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/phpseclib/+bug/1574058/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
