[Yum] Determining best arch in HeaderInfoNevralLoad + SIGNAL stuff

[David Farning]

> This is kinda moot, of course b/c this part is disappearing.
> 
> Have you looked at yum-HEAD/rpmUtils/updates.py and arch.py?
> 
> That's where the new arch stuff is and I think the biarch support there
> is A LOT more exacting.

I have not looked at looked at you new stuff yet.  I picked a cvs tree
from a few weeks ago as my target to study.  Trying to optimize changing
source in a new lang was pretty challenging.  So, I decided to just pick
a release a work with it. I'll take a look tonight--sounds like you have
made some pretty significant changes.


Was looking at the SIGNAL trap issues. In 

http://www.python.org/doc/current/lib/module-signal.html

Bullet point 2 states that there is no signal block in python.

While bullet point 3 two it states that external c functions are handled
as atomic units.  Thus a signal sent during function execution would not
be handled until after the function as returned.  

Also looking at the rpm source, a KILL SIGNAL trap is used before
manipulating the rpmdb.

Thus, the function rpmlib calls (via the rpm-python) should be safe from
being killed at an inopportune moment...

If the above is true, I don't under stand why I thought I mangled my
rpmdb last week while test ctrl-c kills 

Thanks
David Farning