Re: how do you want your britney commits (unfinished)
Hi Niels, On 01-10-17 08:36, Niels Thykier wrote: >> Do you already have an idea in mind how this could/should be >> accomplished? While going through britney, I see several (most, if not >> all, of them hacky) ways this can be done. >> > > At the moment, no, I do not have a brilliant idea. Good to know. This eases my mind a bit. > Short term/for testing purposes, I would just add a comment in the > excuses and then leave the aging as it is. If I read you correctly, I think I would be nearly so far now that we could start doing this once debci is enhanced to listen to britney and provide feedback (hi Antonio). We could use the temporary failures instead of the currently used hard failures or the to be implemented method. Have you already looked at the current state? If so, are there already comments you like to share? Paul signature.asc Description: OpenPGP digital signature
Re: how do you want your britney commits (unfinished)
Paul Gevers: > Hi Niels, > Hi, > On 12-09-17 22:15, Niels Thykier wrote: >> Paul Gevers: >>> - adjust britney2 to not block on autopkgtest failure but instead adjust >>> the required age depending on the results (lower for pass, raise for >>> regression) > > Do you already have an idea in mind how this could/should be > accomplished? While going through britney, I see several (most, if not > all, of them hacky) ways this can be done. > At the moment, no, I do not have a brilliant idea. Short term/for testing purposes, I would just add a comment in the excuses and then leave the aging as it is. > One idea of mine is to create an internal urgency that the age policy > can take to adjust the age. If that is the case, I guess I have to make > sure the order of policies is correct. Minor point, but was it your > intent to change to know urgency ages? Or e.g. age to 20 days, which is > no urgency age, when regressions pop up? > > Paul > I think we would prefer modifying the age-requirement over urgency fiddling. Among other because changing the urgency is "exclusive" so we cannot have multiple policies altering the aging in that case. But I do not really have a good way of doing it atm. (either way) Thanks, ~Niels
Re: how do you want your britney commits (unfinished)
Hi Niels, On 12-09-17 22:15, Niels Thykier wrote: > Paul Gevers: >> - adjust britney2 to not block on autopkgtest failure but instead adjust >> the required age depending on the results (lower for pass, raise for >> regression) Do you already have an idea in mind how this could/should be accomplished? While going through britney, I see several (most, if not all, of them hacky) ways this can be done. One idea of mine is to create an internal urgency that the age policy can take to adjust the age. If that is the case, I guess I have to make sure the order of policies is correct. Minor point, but was it your intent to change to know urgency ages? Or e.g. age to 20 days, which is no urgency age, when regressions pop up? Paul signature.asc Description: OpenPGP digital signature
Re: how do you want your britney commits (unfinished)
Hi Niels, On 12-09-17 22:15, Niels Thykier wrote: >> My aim is to implement the support for autopktest in such a way that it >> keeps on supporting the use-case for Ubuntu, while supporting the Debian >> use-case as well. Do you want/appreciate that as well? >> > > Sounds reasonable from a social PoV, though I admit I am unaware of the > technical consequences of that. Though they are presumably minor? So far, I think it will only mean that there is code that will not be used by Debian. And maybe the action of getting the result into britney2 would be located in a different position in the code than where it currently sits. > Feel free to use a branch on github or create a user-specific repo on > alioth. Either way will be fine with me. :) Pushed to github. Paul signature.asc Description: OpenPGP digital signature
Re: how do you want your britney commits (unfinished)
Paul Gevers: > Hi Niels, > Hi > I am (slowly) progressing through my britney2 changes for autopkgtest > support. I have incorporated the Ubuntu implementation on the Debian > britney2 stack and have fixed the regression tests that Ubuntu provided, > at least those that I think should remain working in the Debian setup, > which is the fast majority. > Ok. :) > My aim is to implement the support for autopktest in such a way that it > keeps on supporting the use-case for Ubuntu, while supporting the Debian > use-case as well. Do you want/appreciate that as well? > Sounds reasonable from a social PoV, though I admit I am unaware of the technical consequences of that. Though they are presumably minor? > Now I'd like to already share what I have with you, but it isn't > finished yet. I am working on a dedicated branch for now. Do you want me > to push this branch to github already, or doesn't that work as intended > (by me)? > Feel free to use a branch on github or create a user-specific repo on alioth. Either way will be fine with me. :) > My next steps will be: > - add support for debci results from file (additional to the Ubuntu > implementation where results are directly downloaded by britney2 from swift) > - make URL's in excuses (for autopkgtest results) configurable to enable > code sharing between Debian and Ubuntu (and other derivatives) > - adjust britney2 to not block on autopkgtest failure but instead adjust > the required age depending on the results (lower for pass, raise for > regression) > > Paul > Sounds good. :) Once again, thanks for working on this. :) ~Niels
