Message307843
| Author |
xdegaye |
| Recipients |
vstinner, xdegaye |
| Date |
2017年12月08日.09:42:16 |
| SpamBayes Score |
-1.0 |
| Marked as misclassified |
Yes |
| Message-id |
<1512726136.75.0.213398074469.issue32246@psf.upfronthosting.co.za> |
| In-reply-to |
| Content |
I agree that test_crashed() should be skipped on Android anyway and I will create a PR to skip the test.
If I understand correctly, there may be another nasty bug that is revealed when test_regrstest fails with the environment altered:
How is it possible that the new file that alters the environment be a test_python_* directory (see my initial post) when all the tests are run inside a TEMPDIR/test_python_* directory. The 'files' environment is checked within a TEMPDIR/test_python_* and not within TEMPDIR, no ?
I found a new point that may help understand this problem:
* On Android TEMPDIR is built from tempfile.gettempdir() that uses the TMPDIR environment variable which is set by Android to /data/local/tmp. The tests are run in /data/local/tmp/python [1], this is also the value of $SYS_EXEC_PREFIX and thus where are installed Python machine-specific binaries.
* When TMPDIR is set to /data/local/tmp/python/tmp, which makes more sense here, test_regrstest is ok and does not change the environment.
[1] This is the only location where a shell user may have both write and exec permissions, the Android applications have those permissions here too and in their own dedicated locations on /data/data. |
|
History
|
|---|
| Date |
User |
Action |
Args |
| 2017年12月08日 09:42:16 | xdegaye | set | recipients:
+ xdegaye, vstinner |
| 2017年12月08日 09:42:16 | xdegaye | set | messageid: <1512726136.75.0.213398074469.issue32246@psf.upfronthosting.co.za> |
| 2017年12月08日 09:42:16 | xdegaye | link | issue32246 messages |
| 2017年12月08日 09:42:16 | xdegaye | create |
|