Leonard den Ottolander wrote:
Hello Hubert,
On Mon, 2011-01-03 at 01:33 -0600, Hubert Bahr wrote:
I changed the spec file to allow up to 5 errors and it completed the build.
That sounds ugly. If you do this kind of patching you should patch out the specific test failures, not set a number of random tests that are allowed to fail. Just a thought.
All a matter of perspective. This is one of over 2000 source packages to build binary's and it passes 99.9% of the test cases, and this 5 character change to a specfile allowed it to build. The previous alternative was to kill it during a test and then completing the build. The other one was to skip the testing in total. There are 433 bugs filed against nss. Some were just expired certificates causing problems. I chose not worry about it unless I run into further specific problems that point me back to nss. In this case I just wanted to have a baseline should I have need to modify nss source (unlikely unless it is for branding purposes). If I make an error that increases the count above current value, I would still be alerted. My intuition says it is probably a problem with the test case, not a problem with the source or build technique. I would like to hear from anybody who has built rhel-6 nss from RedHat's source rpms, on their results and what technique they applied.
Hubert
Regards, Leonard.