> Hi, > > some days ago you made a change to downloads.c that should fix a range > mismatch when requesting the last byte of a file. That was quite > amusing, because I was just about to report that bug :-) > > Unfortunately, the error is still there, despite that fix. > > My GTKG from latast anonymous cvs has downloaded a large file from > many hosts simultaneously (always more than 50 active downloads at a > time, from thousands of known IP addresses). That went quite well, with > high download rates, until there were only a few bytes left. > > Now it infinitely tries to get the last bits of that file, and will > likely never succeed. It always requests a 513 byte range at the end of > the file (my overlap is set to 512 bytes, so it seems to actually be a > one byte request), and only gets negative replies. Knowing how well it > went before, it just has to be a buggy request, not a lack of sources. > > BTW, even if you really fix this issue, so that everything is correct - > is it really necessary or helpful to break down the file into chunks as > little as one single byte? Your swarming strategy works quite well for > me if the number of sources corresponds to the file size in a way that > is "well-balanced". But if there are much more sources available, it > seems to become inefficient.
Could you please check/verify if/that the bug is now gone in current CVS version by yesterday's patch? Greetings, Thomas.
pgpfZjNjEapoZ.pgp
Description: PGP signature
