Nils Goroll wrote:
> Hi Peter,
> 
> thank you for having a look at my suggestion.
> 
>> I must admit I dislike the use of the goto's within the suggested fix -
>> I would rather see the missing calls to meta_unlock(local_sp...) added
>> in.
> 
> I had thought about this. The reasons why I thought the gotos were 
> somehow better were that
> 
> - they remove duplicate code and
> - the labels add some semantics.
> 

Yes; understood.

> With these points in mind, do you still want the gotos removed? I think 
> it is important to have a solution which the primary maintainers are 
> happy with, even if it's not the one I'd prefer (honestly and seriously, 
> no bad feelings here).

I just don't like goto's :-) Yes removing the duplicate code is
goodness but this is trivial duplicate code. This is my personal
preference. Indeed in the C-style guide we have:

"While not completely avoidable, use of goto is discouraged."

So I suggest have a look at how the ret from meta_set_purge is dealt
with and then code accordingly.

Once we have a sponsor assigned to this the formal code review can
proceed.

> 
>> With the change on line 1329 you are removing the ability to detect 
>> where in the meta_set_purge code a failure occurred. The meta_set_purge
>> code returns 0 on success and => 1 on failure (have a look how the rval
>> is used within that function).
> 
> Oops, good catch, thank you. I'll change this after your next reply.

Okay.

> 
>>> http://cr.opensolaris.org/~nigoroll/bz12149_sdssc_host.c.patch
>>
>> Ah of course! It is not part of the ON base...apologies - separating 
>> out these is good (ie have the ON ones and then the OHA one) and not 
>> including it in the ON changes.
>>
>> Couple of comments here:
>>
>> o I'd called the variable mde (as per lots of the SVM code)
>> o initialise the variable with mdnullerror:
>>
>>     md_error_t    mde = mdnullerror;
> 
> Thanks, I've changed the diff.

For the OHA stuff the process to integrate is at:

http://hub.opensolaris.org/bin/view/Community+Group+ha-clusters/sponsor-tasks

which has some specific cluster links off it.

Ta
pete
> 
> Nils

Reply via email to