Bug #587
closed
Strange changes in general regression test set
Added by Evgeny Novikov over 13 years ago.
Updated over 13 years ago.
Detected in build:
a5ce32a
Published in build:
876e93b
Description
After some fixes in BCE some its problem descriptions change. For example, on middle regression tests:
3 'Unmatched' and 3 'No Makefile or Kbuild file' becomes 5 'Unmatched' and 1 'No Makefile or Kbuild file'.
I notice that there are such descriptions that lead to 'Unmatched':
- Driver have no kernel modules.
- Incorrect driver archive type.
I add corresponding patterns to the problems script. So soon (tomorrow) there won't be 'Unmatched' for the general test set for all tools (this was just for BCE indeed).
But we have to know why such problems have place. May be some fails of BCE, may be some problems in the test set (incorrect archives, packaging scripts, etc.).
Related issues
1 (1 open — 0 closed)
This bug is related with the bug #584 since I can't see where a particular problem arises (in what driver).
- Category changed from 26 to Tests and QA
The category 'Regression testing' should be deleted. Sorry that I skipped 'LDV tests'.
Details of changes:
- '' -> 'Incorrect archive type'
online-drivers--bce_report_test--garbage_driver.tar.bz2.tar.bz2
online-drivers--some_other--a.c_out.tar.bz2.tar.bz2
- 'No Makefile or Kbuild file' -> 'No modules'
online-drivers--from_oldserv--acpi-acpica-20100408.tar.bz2.tar.bz2
online-drivers--from_oldserv--pnp-pnpacpi-20100408.tar.bz2.tar.bz2
- '' -> 'No modules'
online-drivers--some_other--a.c.tar.bz2.tar.bz2
I updated the general regression task file but we have to find out with what these problems are related (are these a real or artificial problems or may be there are bugs in BCE).
It's a BCE bug.
BCE was changed for work with em803 driver. (This driver includes a KERNEL_LOCATION define that is not overridden.)
1. '' -> 'Incorrect archive type'
Comes with ldv-online fixes. (It's a correct description for ldv-online users)
2. 'No Makefile or Kbuild file' -> 'No modules'
It is a new correct description.
Previously, Makefile's and Kbuild files were tested for the presence of "obj-m" declaration. 'No Makefile or Kbuild file' printed when driver Makefile or Kbuild not contains "obj-m". But this description is confusing to online users. In fact, the file exists, but the declaration "obj-m" is not there. For this reason "obj-m" test was removed from BCE and new correct message was printed.
3. '' -> 'No modules'
Driver has correct Kbuild with incorrect source file. Kernel compiled this driver with 'No modules' message.
- Status changed from New to Open
- Assignee set to Vadim Mutilin
So changes are intended and positive but what with sources? I think that Vadim can answer this question.
- Tracker changed from Bug to Feature
- Subject changed from Strange changes in general regression test set to Analyze BCE errors from middle regression test set
- Assignee changed from Vadim Mutilin to Evgeny Novikov
There is 63 build problems of total middle tests 305 launches (about 20%). Just 2 of them seems artificial ones. I think that we really need to update middle tests since 61 drivers aren't checked and just spend our time.
- Tracker changed from Feature to Bug
- Subject changed from Analyze BCE errors from middle regression test set to Strange changes in general regression test set
- Status changed from Open to Closed
- Assignee changed from Evgeny Novikov to Alexandr Strakh
- Published in build set to 876e93b
Sorry for incorrect issue. Of course this must be moved to the separate one!
Also available in: Atom
PDF