Bug#978749: requests: circular dependency makes requests unbuildable
On 12/31/20 4:31 PM, Daniele Tricoli wrote: > Nowadays for > urllib3 upstream uses a range of versions, and I take care of both urllib3 and > requests, so we should not have problems. Oh, it's very nice that upstream changed their mind, after all of this time. Thanks for sharing this info. I can clearly remember heated discussions in the OpenStack list where they really stand on the opinion they should vendor everything and that everyone suggesting otherwise was a fool. :) > Thomas thanks for spotting the missing "nodoc", I will fix also this but not > with the upload that will close this bug. Great! :) > Many thanks, cheers and happy new year! Happy new year too. Cheers, Thomas Goirand (zigo)
Bug#978749: requests: circular dependency makes requests unbuildable
On 2020-12-31 16:03:48 +0100, Thomas Goirand wrote: > On 12/31/20 1:25 PM, Sebastian Ramacher wrote: > > Hi Thomas > > > > On 2020-12-31 13:06:37 +0100, Thomas Goirand wrote: > >> Hi Sebastian, > >> > >> I'm challenging the fact that a circular dependency is a problem for > >> arch:all packages. At least, it shouldn't be an RC bug, IMO. > > > > If packages cannot be built on the buildds, they are RC-buggy. > > Which is a different topic. > > >> I also don't understand why you've renamed the original bug, which seems > >> to be very different from the circular dependency you're describing. Can > >> you explain why they are related, and why we shouldn't have 2 bugs? > > > > The bug is that a build of requests is missing. The version that fixes > > the uninstallability of python3-requests was already uploaded. > > If I understand correctly, the issue was that requests had: > > python3-chardet (>= 3.0.2), python3-chardet (<< 3.1.0) No, the issue was that requests with a fixed dependency on python3-chardet was not buildable because it build-depends on sphinx which in turn depends on python3-requests which was not installable. Cheers > > as (build-)depends, which prevents using charted 4. This is very > different from a circular dependency problem. > > Cheers, > > Thomas Goirand (zigo) -- Sebastian Ramacher signature.asc Description: PGP signature
Bug#978749: requests: circular dependency makes requests unbuildable
Hello Thomas, hello Sebastian! Thomas, thanks for always caring about requests! On Thu, Dec 31, 2020 at 04:03:48PM +0100, Thomas Goirand wrote: > If I understand correctly, the issue was that requests had: > > python3-chardet (>= 3.0.2), python3-chardet (<< 3.1.0) > > as (build-)depends, which prevents using charted 4. This is very > different from a circular dependency problem. As discussed on IRC I'm going to drop the upper bounds for both chardet and urllib3: they was introduced at the time when requests was vendoring all its dependencies and it was more coupled especially with urllib3. Nowadays for urllib3 upstream uses a range of versions, and I take care of both urllib3 and requests, so we should not have problems. chardet is more mature and should not be a problem at all. For the severity of this issue, well I think that we all want Debian in the best shape, so I don't care too much, it's something to be fixed! :) Thomas thanks for spotting the missing "nodoc", I will fix also this but not with the upload that will close this bug. Many thanks, cheers and happy new year! -- Daniele Tricoli 'eriol' https://mornie.org signature.asc Description: PGP signature
Bug#978749: requests: circular dependency makes requests unbuildable
On 12/31/20 1:25 PM, Sebastian Ramacher wrote: > Hi Thomas > > On 2020-12-31 13:06:37 +0100, Thomas Goirand wrote: >> Hi Sebastian, >> >> I'm challenging the fact that a circular dependency is a problem for >> arch:all packages. At least, it shouldn't be an RC bug, IMO. > > If packages cannot be built on the buildds, they are RC-buggy. Which is a different topic. >> I also don't understand why you've renamed the original bug, which seems >> to be very different from the circular dependency you're describing. Can >> you explain why they are related, and why we shouldn't have 2 bugs? > > The bug is that a build of requests is missing. The version that fixes > the uninstallability of python3-requests was already uploaded. If I understand correctly, the issue was that requests had: python3-chardet (>= 3.0.2), python3-chardet (<< 3.1.0) as (build-)depends, which prevents using charted 4. This is very different from a circular dependency problem. Cheers, Thomas Goirand (zigo)
Bug#978749: requests: circular dependency makes requests unbuildable
Hi Thomas On 2020-12-31 13:06:37 +0100, Thomas Goirand wrote: > Hi Sebastian, > > I'm challenging the fact that a circular dependency is a problem for > arch:all packages. At least, it shouldn't be an RC bug, IMO. If packages cannot be built on the buildds, they are RC-buggy. > > If the problem is bootstraping (which isn't a problem, since that's > arch:all), then a wishlist bug against requests so that it implements > the "nodoc" build profile is probably a better way to go. > > Please downgrade the severity. That would be wrong. See https://release.debian.org/bullseye/rc_policy.txt > I also don't understand why you've renamed the original bug, which seems > to be very different from the circular dependency you're describing. Can > you explain why they are related, and why we shouldn't have 2 bugs? The bug is that a build of requests is missing. The version that fixes the uninstallability of python3-requests was already uploaded. Cheers > > Cheers, > > Thomas Goirand (zigo) -- Sebastian Ramacher signature.asc Description: PGP signature
Bug#978749: requests: circular dependency makes requests unbuildable
Hi Sebastian, I'm challenging the fact that a circular dependency is a problem for arch:all packages. At least, it shouldn't be an RC bug, IMO. If the problem is bootstraping (which isn't a problem, since that's arch:all), then a wishlist bug against requests so that it implements the "nodoc" build profile is probably a better way to go. Please downgrade the severity. I also don't understand why you've renamed the original bug, which seems to be very different from the circular dependency you're describing. Can you explain why they are related, and why we shouldn't have 2 bugs? Cheers, Thomas Goirand (zigo)