Repository navigation
v0.20.1
Fixed
-
resolve_kernelno longer returns a different filename than the one asked for (#9).find_by_filenamejoins onsha256, so two byte-identical files under different names share onekernelsrow and either name matches both of theirlocationsrows. The winner was whichever path the scanner walked last, soresolve juice_crema_5_2.tmcould hand backjuice_crema_5_2_v473_20260819_001.tm. A location actually named as requested now always wins.No caller could ever have loaded different data than it asked for: the join is on content, so every path returned was byte-identical to the file requested. What was wrong was the name, which matters for recording which file a run used.
The content match is kept as the documented fallback —
jup365.bspresolving to a location namedjup365_19900101_20500101.bspwhen no file of the requested name is registered — and now adds a warning naming the file used.
Full changelog: https://github.com/michaelaye/spice-kernel-db/blob/main/CHANGELOG.md