On 23/01/18 16:18, Adam Williamson wrote:
On Tue, 2018-01-23 at 08:42 -0700, Jerry James wrote:
Where are the instructions? Why is informing packagers, the group
most affected by this change, an afterthought? We should have been
told about all of this, in detail, prior to the thing being turned on,
*well* before it was turned on.
So, my answer to this is second-hand - apologies if any of it is wrong,
and the folks involved will no doubt correct me. But as I understand
it, I think they weren't really expecting the tests that were turned on
to have false positives; the tests that were chosen were intended to be
ones that wouldn't cause this kind of problem, so there'd be more time
to get all the waiverdb integrations in place. I don't think they
actually anticipated that false positives would happen and people would
need to use waiverdb-cli like this, which is why instructions for it
weren't part of the plan. I'm sure no-one intended to cause disruption
and inconvenience, and we're sorry about it. I'll try to ask the
relevant folks if perhaps we should stop gating on the abidiff test for
now as a short-term measure, or something like that.
There was one that came up on IRC yesterday that had failed abidiff
entirely because of added functions as far as I could see.
I don't know if that counts as a false positive or not because I don't
know what the design goals are - added functions should mean it's not
going to break any existing programs but new programs built against it
might not work against old versions of the library.
Tom
--
Tom Hughes (tom@xxxxxxxxxx)
http://compton.nu/
_______________________________________________
devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
