Re: [MirageOS-devel] Logging (was: test, quality, tcpip)

[Anil Madhavapeddy <anil@xxxxxxxxxx>]

> On 12 May 2015, at 16:46, Richard Mortier <mort@xxxxxxxxxx> wrote:
> 
> On 12 May 2015 at 17:30, Anil Madhavapeddy <anil@xxxxxxxxxx> wrote:
>>> If I understand this correctly, this means that there is one branch for 
>>> errors, one for warnings etc. I have some doubts about this since messages 
>>> occuring in a given level may be due to a message occuring in another level 
>>> (you get a warning because there was an error). So it seems preferable to 
>>> have the linear occurences of each level in the same branch. I would rather 
>>> see branches being used for delineating components.
>> 
>> Agreed. I had in mind a branching structure that would include a branch per 
>> component, followed by the debug levels.  Other listeners could then do 
>> aggregation into different views, including operations such as time-based 
>> rotation and aggregation across components.  For an Irmin backend, such 
>> listeners can just be watchers (either offline or online, which is nice!).
>> 
> 
> Sorry, may be being slow but if I understand you both right this means:
> 
> + one Irmin log ("repo"?)
> + each component has its own branch within that log
> 
> then either
> 
> + within each component branch, there is a linear sequence of messages
> at different levels

This makes sense to me.

> 
> or
> 
> + within each component branch, there is a hierarchy of subordinate
> branches, branching each time a message appears at a lower (finer
> grained) debug level

And this doesn't :)

-anil
_______________________________________________
MirageOS-devel mailing list
MirageOS-devel@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/cgi-bin/mailman/listinfo/mirageos-devel