The "to do" size of the queue is not reduced by any search criteria.
Add template variable noofslots_total to reflect the total number of un-paused items.
- Config support updated to save categories by server. Any server without categories defaults to the Default category.
- Downloader now checks categories before selecting a server to use.
- Servers with the Default category are always eligible.
- Otherwise, only servers with a matching cagegory are eligible.
- Servers without categories are not used (this can happen if you delete the last category for a server).
- Properly skip servers with higher priority if they have no matching category
- Upgrades from previous versions should retain the expected behavior that all servers are eligible until configured otherwise.
Core code changes donated by user Fmstrat.
The current implementation is bit rough as the calculation of alternative servers isn't done at the right level in the download/decoding chain.
This needs a redesign, although it works.
When an explicit priority is set, the category evaluation should not temporarily change the priority,
which will cause a re-sort within the priority group.
Use "\\?\" prefix for all path-related handling in post-processing.
Use shortened paths for unpacking and user scripting.
Also support oversized paths in "incomplete" as long as the base folder is below 260.
The generic version is based on the md5sum of the combined article-identifiers and no longer on the NZB backup.
Episode detection is based on the name/season/episode extraction from the job title.
There's now an option for generic detection and an option for episode detection.
All info is stored in the History database, under a new user_version number "1" because two new columns are needed.
After scheduler evaluation, only set/reset job-pause after considering the current status.
When adding new job to queue: pause job when schedule evaluates pause, but don't undo an already set pause.
When changing priority: only pause when job is checking/downloading/queued, only resume when it was paused.
Don't test for "isinstance(x, str) or isinstance(x, unicode)" but for "isinstance(x, basestring)".
Correct some places that forgot to test for unicode (so test for basestring).
SFV verification per PAR-set using only the matching SFV file.
When no par2 files are found, use all available SFV files.
Remember the verification status of each set in the "verified" marker file.
Improve par-set matcher, so that there's no mix-up when one set name
is a substring of another set name.
When at startup, there's a job in the download which has no more file to download,
it will be passed to post-processing. However the queue logic will try do disconnect
all servers first. This call will fail because the downloader hasn't start yet.
The OSX queue menu entry shows a maximum of 10 jobs.
However, the counter should show the total amount of active jobs along with
the total size of these jobs.
Instead only the contribution of the 10 visible jobs was shown.
Previously a pre-checked job re-entered at the end of the queue.
Moved the quality check from the post-processor to nzbqueue, so a successful job doesn't need to go through the post-processor.
Improvement: all other ways for a job to be sent to post-processing need to go through the end_job() method.
Basic problem: prevent par2 files that have already been downloaded from re-entering the queue.
The issue was introduced after removing the habit of 0.6.x to download a part of each par2 file.
The advantage of the 0.6.x method was that more was known about the par2 files,
the disadvantage that quite some data was downloaded that was never used.
Add job flag that prevents further QuickCheck attempts on known to be damaged jobs.
The method of inspecting large par2 files is time-consuming if nothing is ever found.
The Windows Tray icon queried for free diskspace, while it is never shown.
The OSX top menu should query for no more queue entries than it will show (10).
The OSX top menu queue didn't show accented characters properly.
Jonathon Saine
shypike
cdheiser