DescriptionPetr Viktorin (pviktori)
2022-07-20 12:52:34 UTC
A test for unpacking tarballs with sparse files fails on ppc64le on Koschei (Rawhide & f36):
FAIL: test_sparse_file_00 (test.test_tarfile.GNUReadTest)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/builddir/build/BUILD/Python-3.6.15/Lib/test/test_tarfile.py", line 971, in test_sparse_file_00
self._test_sparse_file("gnu/sparse-0.0")
File "/builddir/build/BUILD/Python-3.6.15/Lib/test/test_tarfile.py", line 965, in _test_sparse_file
self.assertLess(s.st_blocks * 512, s.st_size)
AssertionError: 131072 not less than 86016
Note that 131072 is 256*512.
The issue might be in *reporting* allocated blocks, anywhere along the path:
- Python's assumption that st_block is measured in 512b units (AFAICS it should cuse DEV_BSIZE to be portable)
- converting stat.st_block to the Python stat result object
- libc stat()
- kernel/filesystem
- the test logic using some other wrong assumption
Or the issue might be in the actual tarball unpacking.
> This was fixed in RHEL 8 but not in Fedora :(
Oh sorry, I only fixed the issue in Python upstream and in RHEL. The issue is very specific to test_tarfile on ppc64 with tmpfs filesystem. It's good to backport the fix, it's simple and should repair Koschei :-)
A test for unpacking tarballs with sparse files fails on ppc64le on Koschei (Rawhide & f36): FAIL: test_sparse_file_00 (test.test_tarfile.GNUReadTest) ---------------------------------------------------------------------- Traceback (most recent call last): File "/builddir/build/BUILD/Python-3.6.15/Lib/test/test_tarfile.py", line 971, in test_sparse_file_00 self._test_sparse_file("gnu/sparse-0.0") File "/builddir/build/BUILD/Python-3.6.15/Lib/test/test_tarfile.py", line 965, in _test_sparse_file self.assertLess(s.st_blocks * 512, s.st_size) AssertionError: 131072 not less than 86016 Note that 131072 is 256*512. The issue might be in *reporting* allocated blocks, anywhere along the path: - Python's assumption that st_block is measured in 512b units (AFAICS it should cuse DEV_BSIZE to be portable) - converting stat.st_block to the Python stat result object - libc stat() - kernel/filesystem - the test logic using some other wrong assumption Or the issue might be in the actual tarball unpacking.