Re: [PATCH 2/3] vmevent: Factor vmevent_match_attr() out of vmevent_match()

On Fri, Oct 12, 2012 at 03:37:43PM +0300, Pekka Enberg wrote: [...]

...

+static bool vmevent_match_attr(struct vmevent_attr *attr, u64 value) +{

•   u32 state = attr->state;
  

•   bool attr_lt = state & VMEVENT_ATTR_STATE_VALUE_LT;
  

•   bool attr_gt = state & VMEVENT_ATTR_STATE_VALUE_GT;
  

•   bool attr_eq = state & VMEVENT_ATTR_STATE_VALUE_EQ;
  

•   bool edge = state & VMEVENT_ATTR_STATE_EDGE_TRIGGER;
  

•   u32 was_lt_mask = VMEVENT_ATTR_STATE_VALUE_WAS_LT;
  

•   u32 was_gt_mask = VMEVENT_ATTR_STATE_VALUE_WAS_GT;
  

•   bool lt = value < attr->value;
  

•   bool gt = value > attr->value;
  

•   bool eq = value == attr->value;
  

•   bool was_lt = state & was_lt_mask;
  

•   bool was_gt = state & was_gt_mask;
  

[snip]

So I merged this patch but vmevent_match_attr() is still too ugly for words. It really could use some serious cleanups.

Thanks a lot for merging these cleanups!

Yes, the patch wasn't meant to simplify the matching logic, but just to let us use the function in other places.

I once started converting the function into table-based approach, but the code started growing, and I abandoned the idea for now. I might resume the work just for the fun of it, but the code will be larger than this ad-hoc function, althouh surely it will be more generic and understandable.

But let's solve primary problems with the vmevent first. :-)

Thanks, Anton.