On 11/18/2016 11:05 AM, Piotr Ożarowski wrote: > [Rene Engelhard, 2016-11-18] >>> [...] >>> To sum up: it's IMO really too late for such an upload, and it shouldn't >>> be allowed, unless we don't care about OpenStack in Debian. >> >> Yup. Transition freeze was ~2 weeks ago, and this wasn't approved, so I've >> no idea why this should be allowed... > > that's because: > a) I consider SQLAlchemy new upstream releases stable now, so I dropped > the << dependency in latest upload > b) I wanted to upload it over a month ago. The << 1.1 dependency is > mostly in OpenStack packages so I contacted Thomas back then. He > asked me to wait 1 or 2 weeks because he's busy.
Piotr, I urge you to stop distorting reality. I've told you I needed 2 weeks to check if that would work, and the result is: it does *NOT*. SQLA 1.1 breaks at least Cinder, Barbican and SQLA-Migrate at build time, and for sure even more at runtime. > I agreed to upload to > experimental for now so that he can do tests and asked him to not complain > later about library transition freeze (sic!). He didn't report back any > problems except SQLAlchemy Migrate (about which he wanted to talk > with SQLAlchemy's upstream author during OpenStack conference IIRC). This is *NOT* true. This document has been made a few days after you asked me to test: https://etherpad.openstack.org/p/sqla-1.1-transition This clearly shows trace dumps for Migrate, Cinder, and Barbican. > I wanted to upload it later so I contacted him again, no response > until yesterday when he told me I can go to hell basically. I didn't write things this way at all. I told you SQLA 1.1 breaks things, and that I do not approve the upload. I have also asked you relevant questions on IRC, like why was it so important to upload a new major release so close from the release. Anyway, you have *NEVER* used the normal Debian procedure to do library transitions. This isn't the first time SQLAlchemy is completely breaking OpenStack, and apparently, you don't care and just follow *your* own agenda. > Most of packages are ready now, even without rebuild. I will rebuild all > that are not in OpenStack team over the weekend. Rebuilt packages work > fine with 1.0 an 1.1 (rebuild will simply remove << 1.1 dependency) Why would you do all but OpenStack? You just don't care about it? This doesn't fly. Please revert your upload or provide patches for absolutely all packages you're breaking, and NMU welcome too. And while you're at it, you're also welcome to run all functional tests... For such a package as SQLAlchemy, a 2 weeks plan doesn't work. In the OpenStack upstream world, it can take a full development cycle so that everyone switches to a major release. That's what I expect to happen also in Debian, or at least to give enough time to organize things with upstream. I would also eventually suggest that you drop yourself from the maintenance of SQLAlchemy, as you're breaking leaf packages without care. I'm not even scratching the fact you never use the transition mechanism that we have in place in Debian (and it's the 4th time like this, even if I asked you multiple times to use transtion as per the release team process). If the release team doesn't express his strong opinion so that you revert your upload, I may ask the TC as you suggested yesterday after a single line I dropped you on IRC (which I find a *very* aggressive way to engage into a conversation). Cheers, Thomas Goirand (zigo)
