On Mon, Aug 16, 2010 at 01:45:45PM -0500, Jamie Strandboge wrote:
Why do we need special behaviour to skip files which don't exist ? If this is solely to keep the test suite happy, then IMHO the fix should be in the test suite. Runtime production code shouldn't have workarounds for test code.
This is not to work around the test suite.
The intent is to help people migrating from < 0.8.3 to 0.8.3. When considering the following binary states: file exists, disk format is specified, ignoreOpenFailure and allowProbing, what is in trunk (ie ignoreOpenFailure=true) and this patch work the same in all cases except when the file does not exist and the format is not specified and allowProbing=false (the new default). With what is in trunk, this is an error condition, with my patch, the domain will start but the disk is not added to the profile. Throwing an error because the format isn't correct on a disk that doesn't exist seemed an odd side-effect when migrating. I can see arguments against this, and would like to commit the changes to tests/virt-aa-helper-test regardless.
How can the domain start if the configured disk file doesn't exist on the host filesystem ? QEMU will try to open a non-existant file, fail, and abort. Failing on non-existant files when setting up the security profile doesn't change that, it just makes us report the problem to the user soon in the startup process. Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://deltacloud.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|