Large torrents and buggy partfiles

[uecasm]

There's an 80GB torrent file that I'm trying to download. Since I have a download cap, I can't get it all at once; the plan was to select about a third of the files this month, then another third next month, etc.

At first I selected about 5GB of files to download and put the rest on "skip". These were all downloaded successfully and were fully accessible.

Then I expanded the selection to about 11GB of files. It got to about 50% and then failed with a "read error" according to the logs. Trying to restart the torrent led to a similar error within a couple of minutes. Thanks to a tip in another thread, I checked the partfile and discovered it was 2GB.

I deleted it and restarted the torrent. The progress dropped back to 30% (it now thinks it needs to redownload even some parts successfully downloaded in the first set, which reached 100%!), but it was then able to progress to 70% before erroring out again -- once again, at the moment the partfile hits 2GB.

Again, I deleted the partfile, and resumed the torrent. This time it made it to 98% before erroring out again with a 2GB partfile. Deleting and resuming at this point drops progress back to 60%.

I'm thinking about changing fileio.use_partfile to False (since obviously the partfile implementation is buggy), but the description says that this is necessary for partial downloads, and I need to do partial downloads.

Not that this whole thing is helping my download cap all that much. Apparently I have now downloaded 19GB of the 11GB I selected for download.

Representative error:

[2011-01-19 03:21:40]  IO Error:13 line:1926 align:-99 pos:15728033792 count:16384 actual:-99
[2011-01-19 03:21:46]  ReadFile error: part6.zip:11714723840:16384:16384:16

[uecasm]

(Should this be in Found Bugs?)

[JF]

Anyone? I have this problem and it is pretty annoying. As soon as the partfile reaches 2 g, I have to download last percents again...