Re: [libvirt] [PATCH 1/2] Refactor virMutexInit virRWLockInit and virCondInit
On Wed, Jul 23, 2014 at 07:27:24AM +0200, Martin Kletzander wrote:
On Fri, Jul 18, 2014 at 07:30:38AM -0600, Eric Blake wrote:
>On 07/17/2014 10:49 PM, Jincheng Miao wrote:
>>Implement InitInternal functions for Mutex, RWLock and Cond, these functions
>>contain error report using virReportSystemErrorFull, it is controlled by
>>an argument 'quite'.
>>The related macros are Init and InitQuite, they call InitInternal function
>>by passing 'false' or 'true' to quite argument.
>
>After your patch 2, do we really have any callers that use the quiet
>version? This seems like the sort of patch where ALL callers should be
>noisy (especially since the failure is so rare, but also means we are
>quite hosed if it happens). What are the callers that you intend to be
>quiet, and what is the justification for them being quiet?
>
I think there are few callers that error out with VIR_ERROR(), but
most of them just return -1 with no apparent reason. I'm not sure
about few class initializations (virOnceInit) and driver
initializations. Do we want to be loud on that front as well?
Yes, global initializers must always report proper errors - these
are captured & reported on every call even though the initializer
is only run once. So yes, I think everything should be noisy in
this case - there's no legitimate reason with these fnuctions why
you would want to be quiet and/or ignore failure.
Regards,
Daniel
--
|: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org -o- http://virt-manager.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|