#4635 closed enhancement (fixed)
bind-9.9.5
Reported by: | Fernando de Oliveira | Owned by: | |
---|---|---|---|
Priority: | normal | Milestone: | |
Component: | BOOK | Version: | SVN |
Severity: | normal | Keywords: | |
Cc: |
Change History (11)
comment:1 by , 11 years ago
Owner: | changed from | to
---|---|
Status: | new → assigned |
comment:2 by , 11 years ago
comment:3 by , 11 years ago
I used
make -k check
Confirm that most of the time is for the inline tests.
My results:
$ grep "R:PASS" bind-9.9.5.0-make-k-check-2014.02.11-17h25m15s.log | wc -l 189 $ grep "R:FAIL" bind-9.9.5.0-make-k-check-2014.02.11-17h25m15s.log | wc -l 2
Failed tests: attached file.
by , 11 years ago
Attachment: | bind-9.9.5-failed tests-2014.02.11.log added |
---|
Summary of the tests with log of the two that failed
comment:4 by , 11 years ago
tail of the tests:
.. I:System test result summary: I: 2 FAIL I: 58 PASS I: 2 SKIPPED make[3]: *** [test] Error 1 make[3]: Leaving directory `/home/fernando/tmp/paco-build-2014.02.11-17h25m15s/bind-9.9.5/bin/tests/system' make[2]: *** [test] Error 2 make[2]: Leaving directory `/home/fernando/tmp/paco-build-2014.02.11-17h25m15s/bind-9.9.5/bin/tests' make[1]: *** [test-force] Error 1 make[1]: Leaving directory `/home/fernando/tmp/paco-build-2014.02.11-17h25m15s/bind-9.9.5' make: *** [check] Error 2 real 40m52.148s user 1m40.267s sys 0m53.429s
comment:5 by , 11 years ago
The space in the file name of the attachment was a typo.
In unit/README, there is this comment:
"These unit tests for BIND 9 are based on the NetBSD Automated Test Framework release 0.17."
Then:
"configure --with-atf"
This seems not to be a good idea for build/install.
But for the tests, it added one apparently complete test, that ran without error:
Summary for 27 test programs: 120 passed test cases. 0 failed test cases. 0 expected failed test cases. 0 skipped test cases. R:PASS
and now, I have:
$ grep "R:FAIL" bind-9.9.5.0-make-k-check-2014.02.11-22h04m35s.log | wc -l 2 $ grep "R:PASS" bind-9.9.5.0-make-k-check-2014.02.11-22h04m35s.log | wc -l 190
so, 190, instead of 189.
I have spent some time searching about bind tests in the web, but cannot find anything relevant.
comment:6 by , 11 years ago
However, if the NetBSD Automated Test Framework (atf) is used, it seems the other tests are no longer necessary, but cannot see how to disable them. The negative point is that it would need two different configure/make from clean sources as I have recently found for Python 3.
In bind, I just added the switch --with-atf to the normal ones, but it seems to build a whole lot of new programs and install them, surely some some at /home/<user>, at make install.
comment:7 by , 11 years ago
I've looked at this for some time and have narrowed down the problem somewhat. In the setup of the inline tests, the following is happening (with minor edits):
cd /tmp/bind/bind-9.9.5/bin/tests/system/inline/ns3/ /tmp/bind/bind-9.9.5/bin/named/named -m record,size,mctx -T clienttest -c named.conf -d 99 -g -U 4 >named.run 2>&1
If I skip the actual tests, then the shutdown send a TERM signal to the above, but it does not terminate and has to be killed. The process has a procedure that waits for up to 60 seconds and then issues an ABRT signal and fails the test. There are other problems, but I think this is symptomatic of those too.
There are parallel procedures for ns{1,2,4,5,6} and they work fine.
I'm going to take a look at an earlier version to see if the same problem exists. I also will bring this up on the bind mailing list if I can't resolve it.
comment:8 by , 11 years ago
Got the answer from the bind-users mailing list. We need to add --with-randomdev=/dev/urandom to the configure options.
I built this package and it seems to build and install without problem. Simple manual tests work, but the regression tests hang. After two hours I killed it. I still got 159 tests that passed, but I'm uncomfortable with keeping the regression tests.
I'd appreciate it if someone would try this package through the 'make check' to see if the results are the same.
Digging into the 'make check', it is basically doing:
Note: After I wrote the above, I traced the problem to the "inline" test. There are a lot of long delays there. The tests are all bash scripts and I'll work on tracing down the problem on my system, but I won't be able to get to it today.