On 29.09.2016 16:50, John Ferlan wrote:
On 09/29/2016 10:06 AM, Michal Privoznik wrote:
On 15.09.2016 16:35, Michal Privoznik wrote:
Just read the 3/3. I didn't know whether I should laugh or cry. I did both.
Michal Privoznik (3): lock_driver_sanlock: Avoid global driver variable whenever possible m4: Check for sanlock_write_lockspace sanlock: Properly init io_timeout
m4/virt-sanlock.m4 | 14 +++++--- src/locking/lock_driver_sanlock.c | 69 +++++++++++++++++++++++++++++---------- 2 files changed, 60 insertions(+), 23 deletions(-)
Thanks John for ACKing the series. I'm not quite sure whether I can push it now that we are in the freeze. I mean it can be viewed as a bug fix, so I'm inclined to push it. What are your thoughts?
From the aspect of you have a legitimate bug that's causing a feature to not work properly, sure I agree. From the aspect of I've pushed something during a freeze before and was told I shouldn't have, maybe I'm not the best person to ask ;-).
BTW: I understand your point about not wanting to document a specific version since it's possible (I suppose) that the API could be backported to some earlier release (not that we'd ever do that).
Still, I'm reacting to a feature someone may have thought was working that suddenly will cause a failure to start a domain if they don't have the new API, but did have the old API. At least a running guest won't "go missing". IOW: how can we inform the user that the minimum version we expected now changed because of some change in an underlying package we depend on.
Well, I guess they will find out as soon they try to start a domain. If they are running old sanlock (which they shouldn't at all - we advocate for using our virlockd), they'll see a sensible error message: + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", + _("unable to use io_timeout with this version of sanlock"));
All that said - push and say sorry later ;-)
Done :-) Thank you. Michal
John