On Thu, Dec 23, 2010 at 10:00:25PM -0500, Eric Cooper wrote: > On Thu, Dec 23, 2010 at 12:43:05PM -0800, Vagrant Cascadian wrote: > > it seems like approx downloads pdiffs, applies them to the appropriate > > Packages > > or Sources file, and then throws the pdiff away. this causes approx to > > download the pdiffs multiple times if multiple machines request some of the > > same pdiffs. > > Can you send me a log (with approx's $debug variable set to true) > showing this behavior? It's not supposed to happen. Approx does > apply the pdiff files and discard them if it correctly generates an > updated Packages file. But subsequent clients that ask for pdiffs > should be "denied", and forced to request the entire Packages file, > because it's now available locally from the approx cache. So you > shouldn't see the same pdiff being downloaded remotely more than once. > Note that the output from "apt-get update" will still show requests > for pdiffs (actually 3 lines for each one!), but they should show > "Ignored" for subsequent clients.
after trying it just now, this appears to be how it behaves, though all of the machines are fairly close as to which pdiffs need to be downloaded. i'll try some more experimentation to see if i can trigger the issue. in theory, should approx currently handle if: 1. machine A downloads pdiffs from yesterday 2. approx applies them sucessfully 3. time passes and the Packages file approx cached gets out of date 4. machine B needs to downloads pdiffs from the last 3 days does approx deny machine B the Packages.diff/Index, even though approx doesn't have a current Packages file? how does machine B end up a current Packages file? live well, vagrant -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
