A pack abandoned at import is invisible in the UI #227
Notifications
Due Date
No due date set.
Depends on
Reference: yolo/arr#227
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
When a pack hard-fails at import, the download stays complete in Transmission
and nothing in the UI says the grab was abandoned.
The sequence, from production on Rick and Morty S08:
§5.7 condemned the whole release
hard_fail_tvblacklisted it, set the grab tofailed, and wrote all tenepisodes back to
missingThe result the operator sees is a torrent sitting at 100% in Transmission and a
season reading 0/10 in arr, with nothing connecting the two. They found out by
opening Transmission and asking why.
Every fact needed is already recorded.
grabsholds the failed row, theblacklist holds the release name and the reason string
hard_fail_tvwascalled with —
"no file matches a wanted episode", or the name of the rule thatrejected it. None of it reaches the UI.
Surface it where the operator already looks:
this season failed at import and why, rather than leaving the season looking
as though nothing was ever tried
blacklisted; it should beable to show the reason it was blacklisted for, since that is the difference
between "this release is bad" and "your policy rejected it"
The second one matters most: a size rejection means the operator can waive it
(#210) and try again, while a corrupt release means they should not. Those look
identical today.
Depends on: #211
Non-goals: do not make arr delete or reclaim the orphaned torrent — §7.3 is
deliberate and the reaper owns it. Do not import a torrent arr did not grab;
§2 rules out filesystem scanning permanently. This issue is about telling the
operator what already happened, not changing what happens.
Out-of-scope finding while working this issue, filed here rather than fixed silently.
web/scripts/check-tokens.mjsstrips/* */and//comments before scanning for literal colours, but not HTML comments. An issue reference of three or more digits inside anindex.htmlcomment —<!-- #227: ... -->— reads as a hex colour and failsjust ci. Existing comments only reference two-digit issues, so nothing had tripped it before. Worked around by writingissue 227instead; the checker should strip<!-- -->the way it strips the other two.