On Mon, Dec 21, 2015 at 12:48:45PM +0100, Stefan Fritsch wrote: > apt in jessie does not have the same problem. Have there been changes in > that area? Yeah, and the cause is indeed apt-file, but not in the way you would think: > OTOH, is is also possible the perceived slow-down is due to apt-file now > using apt's transport mechanisms (and apt-file's pdiffs being a lot larger > than apt-get's). Nope as the Contents files are compressed, which as Julian already mentioned doesn't have this problem. It is still apt-files "fault" in sofar as I had to make rred support patching up compressed files by transparently un- and recompressing. Before 1.1 rred could only deal with uncompressed files (the patches were compressed, but the file to patch had to be uncompressed which for Packages and co is [at least by default] the case). I think while writing d7a51997c30b2098bb60b3397095ec58ec825303 I had this 1-char limit in the back of my head as – after all – it was me who made this inefficient and usually not used ReadLine fallback implementation as a "temporary solution" in 2011 [and as usual, nothing is as permanent as a temprory solution], but I forgot about it after getting the feature itself working. After all: it works. Just not in a very optimized way, but better than nothing. :) [I noticed that pdiffs were slower for a while now, but nothing too extreme to make investigating this a high priority. I discarded it as a form of (negative) confirmation bias after talking again and again in defense of pdiff in fact] So, yeah, problem well understood – no need to further debug/check that. "Just" needs some heavy code juggling now to make it fly again… Best regards David Kalnischkies
Attachment:
signature.asc
Description: PGP signature