Conversation
Package Evaluation Report for GAP
|
|
I asked Claude to look into the failures here, these are its findings: Why the 48 packages fail in PackageDistro PR #1183Data source: run 29783367069 Two important preliminary findings1. Note also that SmallGrp and PrimGrp both list GAPDoc as a needed package, so anything that 2. How to read the "signal" columnGAP core binds stubs for exactly three names (
Everything else (
Failure categories
The table
Summary countsPackages can fall in more than one bucket, so these don't sum to 48.
Of the 7 GAPDoc cases, 5 involve no group DB at all (alnuth, digraphs, hecke, idrel, By database (packages where that DB is implicated at all — 5 packages involve none):
VerificationAll 48 packages were re-run locally against GAP 4.17dev-45 as i.e. GAPDoc available, no group databases — which isolates exactly the PrimGrp/SmallGrp/ Result: exactly 5 packages pass — alnuth, digraphs, hecke, idrel, liealgdb — the GAPDoc-only Note Two attributions were established by bisection rather than by reading the error:
Suggested next steps
|
145b12c to
20c4293
Compare
Experiment motivated by gap-system/gap#2434