I think in the end it's mostly two jiras that cause the difference:
PHOENIX-5228 - this one is weird since there seem to be so many unnecessary
differences between the branches, and non seems to be canonically right. It's
pretty big so I can perhaps see how this happened.
PHOENIX-4009 - which was just not pushed to 1.3
The rest are mostly minor changes that seem legit. If we can push PHOENIX-4009
to -1.3 as well, one could do a quick run between the branches and make
PHOENIX-5228 the same, without the need to force-push.
The difference in tests-failures are present, though. PartialCommitIT for
example fails on the -1.3 branch only. And no green run of -1.4 or master, yet.
I've done some work, too, to speed the test up. Some runs now finish in just
over 2h now (never below 3h before).Oh got another green -1.5 run ;-)
-- Lars
On Thursday, June 13, 2019, 8:51:00 PM PDT, Andrew Purtell
<[email protected]> wrote:
What I would recommend is take on branch as the canonical branch, I guess the
1.5 one since Lars says it’s testing out green. Find the common ancestor prior
to divergence. For all other branches reset the head to that ancestor and then
pick forward from 1.5 one commit at a time, fixing up for compile issues due
the the different HBase versions. Then make a push to resolve any lingering
test issues. This means all but the 1.5 branch would have history rewritten and
would be force pushed. However in exchange you’ll have common and consistent
history on all.
If this is an acceptable result I have undertaken this kind of janitorial work
and would be pleased to offer my services for clean up duty. The ginsu I mean
git knives for slicing and dicing history are sharp and at the ready. At the
conclusion of the work you the Phoenix community will have only the last step
to perform, which would be to resolve any lingering test issues.
> On Jun 13, 2019, at 7:29 PM, "[email protected]" <[email protected]> wrote:
>
> Hi all,
> we have three active 4.x branches: 4.x-HBase-1.3, 4.x-HBase-1.4,
> 4.x-HBase-1.5.
> It looks like we're lacking _basic_ discipline here. There are patches in
> some branches but not in others, some patches are different between these
> branches for no good reason, different tests fail, etc.
> I have run out of my available time to track these all down. I filed Jiras to
> that extend, but as is we're completely unable to release a coherent 4.15.0
> version of Phoenix (it will be different between the three versions)
>
> We can either just declare 4.x-HBase-1.3 dead and align -1.4 and -1.5, or
> folks reading this can please push their changes everywhere, or ... I don't
> know we give up? Release anyway?
> The 4.x-HBase-1.5 test suite is passing - I spent a lot of time on at least
> that. There are no excuses like "But it failed before!".(there's at least one
> known flapper, and I'll try to fix or disable that test)
>
> I'll stay to my threat and veto/revert any patch that changes that.If that
> ends up swimming against the stream and eats up too much of my time, I'll
> just give up!
> -- Lars
