* Paolo Bonzini (pbonz...@redhat.com) wrote: > On 12/09/19 19:45, Dr. David Alan Gilbert wrote: > > Do you think it's best to use the block version for all cases > > or to use the non-block version by taste? > > The block version is quite nice, but that turns most of the patches > > into 'indent everything'. > > I don't really know myself.
OK, new version coming up with a mix - the diffs do look a lot more hectic with the block version. > On first glance I didn't like too much the non-block version and thought > it was because our coding standards does not include variables declared > in the middle of a block. I took that as being a coding standard to avoid confusing humans and since it wasn't visible it didn't matter too much. > However, I think what really bothering me is > "AUTO" in the name. What do you think about "RCU_READ_LOCK_GUARD()"? > The block version would have the additional prefix "WITH_". Oh well, if it's just the name we can fix that. > We could also add LOCK_GUARD(lock) and WITH_LOCK_GUARD(lock), using > QemuLockable for polymorphism. I even had patches a while ago (though > they used something like LOCK_GUARD(guard_var, lock). I think we > dropped them because of fear that the API was a bit over-engineered. Probably a separate set. Dave > Paolo -- Dr. David Alan Gilbert / dgilb...@redhat.com / Manchester, UK