On 19. 7. 2026 15:31, Daniel Sahlberg wrote:
Den sön 19 juli 2026 kl 01:18 skrev Timofei Zhakov <[email protected]>:
On Sun, Jul 19, 2026 at 1:07 AM Daniel Sahlberg
<[email protected]> wrote:
Hi,
I'm sure I'm doing something wrong but I can't figure out what.
svn_trunk2 is a regular checkout of trunk.
$ cmake -B out svn_trunk2 -DSVN_ENABLE_TESTS=ON
$ cd out
$ cmake --build
$ ctest
[...]
All Python-based tests (63 and up) fail with the same error
message:
[[[
Traceback (most recent call last):
File "/home/dsg/ramdisk/svn_trunk2/build/run_tests.py", line
1127, in <module>
main()
~~~~^^
File "/home/dsg/ramdisk/svn_trunk2/build/run_tests.py", line
1120, in main
failed = th.run(programs)
File "/home/dsg/ramdisk/svn_trunk2/build/run_tests.py", line
597, in run
self._init_py_tests(list(py_basedir)[0])
~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^
File "/home/dsg/ramdisk/svn_trunk2/build/run_tests.py", line
330, in _init_py_tests
svntest = importlib.import_module('svntest')
File "/usr/lib/python3.14/importlib/__init__.py", line 88,
in import_module
return _bootstrap._gcd_import(name[level:], package, level)
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "<frozen importlib._bootstrap>", line 1406, in _gcd_import
File "<frozen importlib._bootstrap>", line 1371, in
_find_and_load
File "<frozen importlib._bootstrap>", line 1335, in
_find_and_load_unlocked
ModuleNotFoundError: No module named 'svntest'
]]]
(I have also tried with the "out" directory under svn_trunk2
but that had the same problem).
I think it's the regression from here:
https://lists.apache.org/thread/3s8d1t8r6x5d8xvvmqfqsfcm6fs1o56t
The rest - I think you do it right.
Ah, yes of course. I didn't connect the mental dots. Thanks!
I see Jun already committed this patch in r1936304 this morning, with
this the tests work.
I think that one should also be included in 1.15.0 - not necessarily a
release blocker but close to it - much of the thunder from the CMake
build is lost if the tests are unable to run.
CMake isn't really supported on anything but Windows, so making this a
release blocker would, IMO, require at least a discussion + consensus
here on dev@. I'm not happy that it's being somehow gently ushered in as
a replacement for autotools on those platforms that support autotools.
Suddenly making this failure important (it's not a regression (tm)) is
just way off the charts.
-- Brane
