I've initiated an extensive discussion
about the pros and cons of fixing this bug with Dani, John, Tom
and Ian, yesterday.
The gist of is that, the problem we have in SR2 is worst than the
problem we had in SR0/SR1 because it can lead to completely broken
installs whereas in SR0/SR1 the installation would not proceed.
However in SR0/SR1 the support for p2f was completely broken which
means that if it was not fixed in SR2, then the import from file
facility would just be completely broken in Kepler, which itself
is a problem for some users.
In the end we have to do something, and I believe that it is best
to fix the p2 code to have the import wizard work.
As for the safety of the fix, the fix is in the p2 resolver which
is one of the p2 area that has the highest test coverage so I'm
confident. Also I know from past experience doing changes in the
resolver that these tests really help catch regressions.
HTH
Pascal
On 21/02/2014 12:32 PM, David M Williams wrote:
Thanks Pascal. I'll
schedule the re-build
once approved. Sounds like it shouldn't take long.
Can you say more, in the bug,
about
potential risks? Apparently this was caused by a bug fix at the
end of
January, so some might think the safer course could be to simply
revert
the bug fix.
Do you feel you have adequate
test cases
to prove things are correct, once fix is built, or is there a
chance it
too might make some use-cases worse?
[And, don't get me wrong, if
needed,
it's needed ... I'd just like to see more discussion of that
aspect of
the fix -- why better/safer to fix, rather than just revert the
previous
fix?]
Thanks,
From:
Pascal Rapicault
<pascal@xxxxxxxxxxxx>
To:
platform-releng-dev@xxxxxxxxxxx,
Date:
02/21/2014 12:24 PM
Subject:
[platform-releng-dev]
Kepler rebuild for p2
Sent by:
platform-releng-dev-bounces@xxxxxxxxxxx
Hi,
As mentioned on the cross-project ML and previously agreed
with Dani,
John, Tom and Ian, p2 is requesting a rebuild for Kepler SR2
to address
the bug https://bugs.eclipse.org/bugs/show_bug.cgi?id=428575.
We understand that to get this rebuild consumed by the release
train an
exception we will have to obtained from the Planning Council,
and we
will send an email shortly to the PC ML. However we think that
if the
approval goes through then we should have a build ready to go
to not
delay the release of SR2, which is why I'm preemptively
requesting this
rebuild.
Pascal
_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev
_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev
|