Hi,

SQLAlchemy 1.4.47+ds1-1 in unstable is affected by this upstream bug:
https://github.com/sqlalchemy/sqlalchemy/issues/9819

Latest upstream release for 1.4.x (ie: 1.4.50). I tried building it locally,
and rebuild affected packages (for OpenStack), and this fixes these bugs (or
at least, it's part of solving these bugs as there are also other issues
unrelated):

https://bugs.debian.org/1058183 (barbican)
https://bugs.debian.org/1058123 (neutron-vpnaas) (FTBFS reported fixed by eventlet, but package affected by this bug)
https://bugs.debian.org/1058219 (senlin)
https://bugs.debian.org/1058233 (sahara)
https://bugs.debian.org/1058531 (heat) (some other failures, but also affected)

There's probably more (I didn't attempt to rebuild all affected packages yet),
but these are 5 examples of packages that FTBFS because of this problem.

It'd be nice if we could have the patch in Unstable with this series 1.4.x
before uploading 2.x, so that I have a chance to fix all Python 3.12 bugs and
not mix SQLA 2.x and Py 3.12 problems.

Cheers,

Thomas Goirand (zigo)

Hi,

we have quite some Python3.12 related bugs caused by sqlalchemy which
seem to be fixed in experimental (which is lagging behind upstream
2.0.27 as well as version 1.4 in unstable where upstream just released
1.4.51).

It seems the issue that leads to bug #1058265

  >   File "/usr/lib/python3/dist-packages/sqlalchemy/sql/sqltypes.py", line 2061, in Interval
  >     epoch = dt.datetime.utcfromtimestamp(0)

is not fixed in 1.4.51 so sticking to 1.4 maintenance releases will not
simply solve our Python3.12 issues if we do not fix these ourselves with
patches.  So the question remains for Thomas wo wrote in bug #1058895:

 > It'd be nice if we could have the patch in Unstable with this series 1.4.x
 > before uploading 2.x, so that I have a chance to fix all Python 3.12 bugs and
 > not mix SQLA 2.x and Py 3.12 problems.

Do yuo really want to fix 1.4.x (maybe x=51) first or do we want to
switch to 2.0.27 and fix remaining problems with this latest version?  I
have no resources to run all the tests but I stumbled upon this question
when trying to do some general Python3.12 bug hunting.

Kind regards
    Andreas.

Source: sqlalchemy
Source-Version: 2.0.30+ds1-1

I think this bug has been overtaken by events.

  sqlalchemy | 2.0.32+ds1-1  | testing        | source

All the OpenStack packages mentioned in the original bug report now have
either built cleanly recently or have been removed for other reasons.
There doesn't seem much point in keeping this bug open now.

Thanks,