On Thu, Oct 30, 2008 at 2:17 AM, Hugo Doria <hugodoria@gmail.com> wrote:
I have nothing to add to what Nagy said there. He gave the link to a discussion on the ML about how to implement this differently, in a better way, which would also fix this bug. But I have currently neither the time nor the motivation to finish this discussion, and even less to implement it.
On the other hand, we may completely rework -Qu (atm the patch is waiting for Dan's response), which would completely eliminate this problem (and the whole download size feature(?)): http://www.archlinux.org/pipermail/pacman-dev/2008-October/013026.html However, I agree that we should change our compute_download_size method. Maybe we should compute download size during adding the sync package into the target list (and maybe introduce a new helper function for this). This method and the current one have a little drawback. Sometimes we compute download_size even if it is not needed: -Sp. So compute "on request" is may be better, I don't know. (But how to indicate not computed, set -1 in each alpm_pkg_new? Ugly.) Bye