Description of Problem: SSIA, In testing the HPzx1 products, I saw I/O errors
from md5sum during the CDROM test script. The test itself was marked clearly in
_red_ and said "FAILED" yet the overall driver script "redhatready" created a
results.package file. The really strange thing is that I saw this on a Longs
Peak (the SCSI-based system with only CDROM being the _only_ IDE device).
There's no way certification tests would pass with failing information, so it
seems the creation of /redhatready/results.package is a mistake.
Version-Release number of selected component (if applicable):
# rpm -q rhr-core rhr-interact rhr-auto
How Reproducible: uncertain, saw it once and reported it
Steps to Reproduce: here's a scenario where you might reproduce it
1. Use a patched md5sum that can optionally behave as if it received an I/O
error from a read() or write()
2. Run the CDROM tests with the instrumented md5sum binary set to fail 4-5 times
during a test run
3. Observe /redhatready/results.package exists -- yet the tests should report
failure on the console
Actual Results: Mixed messages (Failed but created results.package) about the
status of the tests
Expected Results: Consistent messages - create results.package only if all tests
Additional Information: This is the version 1.8 tests recently posted on the
website -- bugzilla doesn't offer version 1.8 as a choice yet...
The creation of the results.package is not an indication that all tests have
passed. It is created at the end of the test run in all circumstances. It is up
to the reviewer to determine pass/fail status. This is not considered a bug.