[SCM] Samba Shared Repository - branch master updated

Stefan (metze) Metzmacher metze at samba.org
Mon May 21 00:46:47 MDT 2012


>>> Those patches apparently created flaky builds for Andrew
>>> Bartlett, thus I took them back.
>> Unfortunately this is not the correct way to address
>> that issue. I pushed those patches, not you - and if
>> there were problems with them we should have worked
>> together to find the problem and *collaboratively*
>> fixed them.
> Again I have to apologize for posting those patches to a
> publically accessible branch before I had made sure they
> survive autobuild fine. I had asked for comments, not for
> them getting pushed. Sorry that I was not making that clear
> enough in my request for comments/review.

Jeremy, I think you should not push patches, when other's
don't ask you to do so. Please coordinate with the author
if you should push them. That' why I started to add highly visible
'TODO' markers into my commit message, until I'm 100% sure
something is ready for master.

>> I'd like to propose that all reverts must be
>> reviewed by one other Team member other than
>> the one proposing the revert. It's a code
>> change at that point, and should be treated
>> as such.
> I will be fine with it if comes as part of the general rule
> I proposed a while ago that you can't push your own patches.
> Making an exception for a revert does not make sense.
> Technically, I did not do reverts not to blow up commit
> count unnecessarily, I created a new patch that undoes the
> changes. This code was not ready as I posted it, and it did
> cause severe problems for Andrew. So the only sane thing was
> to pull them as quickly as autobuild would allow it. This
> way Andrew was no longer blocked in his work.

[310/1511 in 16m35s] samba3.base.lock(s3dc)
Testing 32 bit offset rangesUnable to setup \locktest -
UNEXPECTED(error): samba3.base.lock.LOCK3(s3dc)
REASON: _StringException: _StringException: Unknown error/failure.
Missing torture_fail() or torture_assert_*() call?


This really indicates that the share_mode patches are the cause,
I also got a similar failure, I don't think it was the same test, but also

The standard would be to add a specific test to selftest/flapping,
but in this case there was (I guess) no specific test.

I think it's ok to revert the patches and fix them offline
(maybe doing private autobuilds in a loop over a whole day,
 to see if the problem is finally fixed) in such a situation.

At the time Volker asked me to cancel my autobuild (which added tests on top
of the share mode patches), the revert was already in autobuild, but I
had not problems
with it as I read the related discussion on the list.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 262 bytes
Desc: OpenPGP digital signature
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20120521/c2c771dd/attachment.pgp>

More information about the samba-technical mailing list