[Yum-devel] yum stuff and urlgrabber questions

[seth vidal]

Hi all
 first and foremost - paging mr tomayko - where are you? michael and I
were just wondering if you still lived and if you still loved us. :)

ok next things:
I wrote the start of the new depsolver routines - most of it is taking
the depcheck from nevral from the old yum and moving the routines over
to the new metadata and gets.

I added the get() function for repos to get files - not completely
finished yet but it works properly for header gets via ranges - one snag
- if you don't have failure_callback defined as _something_ in your call
to the mirrorgroup object then urlgrabber traces back - since it is
listed as an optional argument it should probably not do that.

right now I've done tests on lots of the little places I needed to test
and I'm feeling pretty good about a lot of misc things.

I want to get yum back to the place where it can perform the rudimentary
functions again. Then go back and start adding clean ups and speed ups.
If anyone is bored and looking for some stuff to hack on let me know - I
can come up with some routines I need and somethings I know I left as
FIXME's.

the metadata parsing is faster and lighter than it ever was before.
Thanks to xmlReader and some mucking about the memory footprint of the
metadata import dropped from 120ishMB to 26MB for the primary metadata.
This is a seriously good thing. The speed is way down as well, on my
meager laptop hardware it's taking 13s to import the primary metadata
for all of fedora core 2 test2. The filelists metadata is the problem
child, though. it's taking much longer than I'd like and I'm not sure,
yet, how to get out of it. It's just a lot of data to traverse and it's
a lot of elements. Right now a perk is that we won't often need to call
it - but it'll be a thorn for users later - so something will have to
give there.

Anyway - that's what I have right now - anyone got anything fun to
report?
-sv