[libvirt] [PATCH v2] remote: Refuse connecting to remote socket

[Eric Blake]

On 08/25/2011 02:04 AM, Michal Privoznik wrote:
>> Then again, remoteError already automatically adds the function name
>> (since it is a macro that includes __func__), so your repeat of the name
>> is redundant.
>>
>> Also, why not list the parsed server, in case it helps the user:
>>
>> remoteError(VIR_ERR_INVALID_ARG,
>>              _("using unix socket and remote server '%s' is not supported"),
>>              conn->uri->server
>>
>> ACK with that fixed.
>>
>
> remoteError(VIR_ERR_INVALID_ARG,..) actually prepend passed message with
> "error: invalid argument in". So I wanted the whole message to make sense :)

I think that's a bug in virterror.c then; most clients of 
VIR_ERR_INVALID_ARG are not expecting to fit into a pre-defined sentence 
like this.  Better would be to fix virterror.c to be a sane message, 
then scrub the few clients that would no longer be worded sanely.

>
> So should I push mine, yours, or completely different version?

Cleaning up existing VIR_ERR_INVALID_ARG oddities is a separate patch, 
and shouldn't interfere with you pushing this patch now.  But I'm not 
sure which version to go with, whether it's better to read well now just 
to be changed later with a cleanup, or to read awkwardly now based on 
the assumption of a coming cleanup.

Anyone else have thoughts on the matter?

-- 
Eric Blake   eblake at redhat.com    +1-801-349-2682
Libvirt virtualization library http://libvirt.org