problems with auto-par of corrupted files

[meBigGuy]

I just downloaded the latest version, hoping it was better than 4.32. But, it isn't much better in regard to corrupted posts.

I have a test post (1 missing and 4 corrupted parts).

Problem 1: It has a missing par2 file, so newsbin never even tries to get the rest of the files. It just hangs trying to get the par2. (with or without par2 pause)

Problem2: If I tell newsbin to load all the files except the par2 it gets the ones that are ok, but it won't automatically assemble the incompletes, (so it can't par2 it and unrar it). In fact, the corrupted files just disappear from the download list, and never show on the file list. So, there is no quick way (like the failed file list in 4.32) to know they failed, and no way to easily select them and "assemble incompletes". In fact, the failed files even get green status indicators on the nzb list (although the status says failed).

I can look for them and select them from the nzb list and say to download them again, and then reselect them and say "assemble incompletes". But I don't understand why this has to be done manually.

I've tried with aggressive assembly mode selected and unselected. And playing with retries didn't seem to help.

Is there any way to automate "assemble incompletes". That one thing makes unattended downloads impractical. You always have to go back and reselect the failed files.

Thanks,

mBG

[meBigGuy]

I found the feature to enable the failed files list. But, I can't select "assemble incompletes" from that list.

[itimpi]

I believe that the auto-assembly of incomplete downloads does not happen unless Newsbin thinks that it has at least 90% of a file. If that condition is not met then you have to take manual action.

[Quade]

I believe that the auto-assembly of incomplete downloads does not happen unless Newsbin thinks that it has at least 90% of a file. If that condition is not met then you have to take manual action.

This. Not only that, unless you have at least one PAR2 downloaded, it'll fail the files to the failed list instead of assembling incomplete. People didn't like parless files assembling because they were useless for the most part. The issue for you is that the pars aren't downloading. Without PARS, Newsbin makes no attempt to assemble or repair.

I found the feature to enable the failed files list. But, I can't select "assemble incompletes" from that list.

Move them back to the download list and assemble them there. Pause if they want to return to the failed list. Unless you can get 90% of the file downloaded, you probably won't be able to repair the files.

IF Newsbin has a par for the set and IF it thinks it has 90% of the files available, it'll assemble and attempt repair after the retries expire. The default retries is 2. I'd suggest not setting it much higher than that.

[meBigGuy]

Problem 1 is a bug. If the main par2 (download_name.par2) is missing, it won't even try for the rest if auto par2 is selected. You can par2 from any par2 file, not just to "top" one.

As for "assemble incomplete":

The process of selecting the failed files (from the failed list), re-adding them to the download list for the directory each is supposed to go to (which is impossible to remember), then trying to find them in the download list so I can select them again and then "assemble incomplete" is totally broken. The download list is grouped, and the files may appear across multiple groups. It's nearly impossible to find them.

And, move to top, etc work weirdly. I can't select a part and move it to top. Seems to only work on groups of parts.
(Is there a way to turn off grouping of parts in the download list?).

What I don't understand is why "assemble incomplete" is not the default. The tool should just download what is there and then see if it can be repaired. If it fails par2, then it fails par2. But you can't know if you don't try. Why put an arbitrary "90%" threshold? The idea that if there is any failure you throw out the whole part and force manual intervention seems totally broken.

The tool has become very cryptic and taken a step backwards (in some respects) since 4.32. In 4.32 I could just select the failed files, select "download incompletes" and the remaider of the files would be downloaded to the directories they were originally targeted for.

You do realize that *almost* all unrepairable downloads are repairable if you "download incomplete".

Reporting this meaningfully is make complex by the fact that the tool behavior *seems* to change from run to run. I have to select features, exit the program, delete the db3's, then restart to try to force consistent behavior.

The last run I tried seemed to do the incompletes assembly by itself, but it took a while. But auto-unrar was deselected.

Thanks for responding, everyone.

mBG

[Quade]

The process of selecting the failed files (from the failed list), re-adding them to the download list for the directory each is supposed to go to (which is impossible to remember), then trying to find them in the download list so I can select them again and then "assemble incomplete" is totally broken. The download list is grouped, and the files may appear across multiple groups. It's nearly impossible to find them.

Not really following you. If you select the item from the failed list, hit Ctrl-Y or enter, it'll go back to the download list with the same path it had set when you added the file to the download list.

What I don't understand is why "assemble incomplete" is not the default. The tool should just download what is there and then see if it can be repaired. If it fails par2, then it fails par2. But you can't know if you don't try. Why put an arbitrary "90%" threshold? The idea that if there is any failure you throw out the whole part and force manual intervention seems totally broken.

Because many people download things without pars and they don't want partial useless files cluttering up their download folder. The compromise is it assembles if it has PAR files.

You do realize that *almost* all unrepairable downloads are repairable if you "download incomplete".

This isn't the case. In fact with DMCA takedowns, the opposite is true. What I need to implement and haven't yet is a 10% failure threshold thing. If more than 10% fails to download abandon the download. The 90% number is because most par sets posted only have 10% par blocks. Meaning repair will only work if you have at least 90% of the file. That's the threshold.


Your problem is you're not getting the initial PAR and that might be a bug. There's code to make one of the other pars download if the primary fails. It might be broken. The way to force a par to download is to click the + , expand the files and pause the files above the pars, then unpause a PAR. How often are you running into this? Just trying to gauge it. I'm not seeing it at all with what I download.

Move to top is for the whole set.

Reporting this meaningfully is make complex by the fact that the tool behavior *seems* to change from run to run. I have to select features, exit the program, delete the db3's, then restart to try to force consistent behavior.

I do wonder if some of your problems are self induced then. None of that is normally necessary. Like the download path thing, I think there's some fundamental misconception about how things work. You expect one thing and it might simply not be how it works.

I'll say it again, the WHOLE problem seems to be getting that first par and everything cascades from that.

Edit: I wasn't clear. I'll be looking into the PAR thing today.

[Quade]

I did find a problem with starting the rest of the pars downloading when the smallest PAR fails to download. I'll have a fix in the next beta.

[Quade]

Ok, I think it's in much better shape now. Next beta, how about trying it and letting me know.

[meBigGuy]

Thanks for reading through my chatter.

I'll give the beta a try tomorrow (I hope).

I like the fact that you capitalize the assembled files. That tells me I need to run par2. (I haven't tried auto par2 and unrar since I somehow got auto assembly to work - not sure what made it start working, but I never set failed files now)

It's interesting that download threads are focused on the chunks of a part, rather than one chunk per part. I like that also since single files load so much faster.

[Quade]

Ideally you won't have to mess with running Quickpar with this change. Specifically, the change I made will make it continue down the list of PAR files downloading them one at a time till it gets a valid PAR.

It'll be in Beta 6 which isn't quite ready yet. I'm in the middle of getting some other stuff in there too.