Great, thank you. We have two options now before stretch code freeze:
- do upload of new upstream (even major, eg. 0.9) version before 2017-02-05 (minus 10 days, better sooner) - maintain 0.8.x as a stable serie in stretch with bugfixes, etc. Newer releases will be uploaded normally to unstable and testing after stretch code freeze (+ stretch-backports eventually) I would personally prefer 2nd option as refactor around vdir seems to be pretty huge and 0.8.4 version is working very well (I am using it since it's release) so I would stick to that as it's well tested and stable. Filip On 2017/01/02 22:55, Christian Geier wrote: > I'll backport those changes onto the 0.8 branch tomorrow. I'll send another > email when it's done. > > Am 2. Januar 2017 22:48:21 MEZ, schrieb Filip Pytloun <[email protected]>: > >Hello, > > > >I checked the changes and as they are done on top of huge refactoring > >(introduce of vdir.py), backporting it back to 0.8.4 seems to be too > >risky so I will workaround this occasional FTBFS by marking the test as > >xfail to avoid autoremoval from stretch. > > > >If we want to fix this issue correctly back in stretch after it's is > >stable, it's possible to do upload of new minor version. > > > >Filip > > > >On 2016/12/19 19:06, Christian Geier wrote: > >> Hi, > >> I could reproduce the issue and have fixed it, see the PR on github > >[1]. > >> Depending on how urgent this is, you can either just take the > >os.sync() > >> commit and apply that to the version currently in Debian (which, > >while > >> brute force, certainly fixes this problem), or you might want to wait > >if > >> we find a more elegant solution. > >> > >> Christian > >> > >> [1] https://github.com/pimutils/khal/pull/543 > >> > >> Quoting Filip Pytloun (2016-12-11 13:16:56) > >> > Hello, > >> > > >> > unfortunately I have no longer access to environment where this was > >> > happening. But if you were able to reproduce the issue and this > >fixed it > >> > for you, I'll apply it and make new release. > >> > > >> > Anyway wouldn't it better to ensure data is written to disk > >directly > >> > during db updates and other operations? Eg. use O_SYNC for safety > >as > >> > these operations doesn't happen so often. > >> > > >> > Filip > >> > > >> > On 2016/12/11 01:14, Christian Geier wrote: > >> > > Hi Filip, > >> > > could you perhaps try to change all those sleep()s to > >`os.sync()`? For > >> > > me it seems to fix the issue. > >> > > > >> > > See [0] for a patch. > >> > > > >> > > If this doesn fix the issue, we obviously need to move the sync > >call out > >> > > of the tests and into the db update. > >> > > > >> > > Best regards, > >> > > Christian > >> > > > >> > > [0] > >https://github.com/pimutils/khal/commit/0b636f7633e86b9e136b06e9965cd3af0e3918f2 > >> > > > > -- > Sent from my phone. Please excuse my brevity.
signature.asc
Description: Digital signature
