Re: [mh] Startup network flood

[Gregg L. <gr...@li...>]

On 1/12/2011 12:43 AM, Matthew Campbell wrote:
> I found the source of the network collision problems I'm experiencing
> when starting up MH.
>
> lib/Insteon_PLM.pm:poll_all() calls $obj->request_status() in a tight loop.
>
> Is there an existing mechanism that could be used to only queue the
> status requests?
>
> I am looking at the loop code in
> code/common/insteon_items_commands.pl:_get_next_linkscan.  This is
> more along the lines I am looking for, but it does not currently work
> on my build.  I suspect inelegant error recovery unearthed by a
> mistake in my mht file.  More details to follow once I figure out why
> sync_all_links is dying.
>
> Back to my question. Does such a queue already exist, or does one need
> to be implemented in poll_all()?

There is not a queue of the sort that you are thinking within that 
portion of the insteon code in trunk.  The code in the insteon branch 
has been completely rewritten to avoid collisions altogether and you 
don't see this effect at all.  Unfortunately, the branch has a different 
set of issues as refactoring has left much of the link/memory management 
inoperable. I've been slowly working on this, but it remains an issue 
for now.

 From my perspective, I've orphaned the trunk code except for minor 
fixes (especially since I can't easily test it anymore).  So, feel free 
to make changes if you want.  But, the downside is those changes might 
not make it into the branch (because the design is so very different) 
which will eventually replace the trunk's insteon implementation.

Gregg