[Loyale]

Fixes all ten findings from a full review of main. Three of them came from #3: the reporting schedule, the notice date and the funding totals. No migrations, and no changes to stored records.

Reports and awards

• Current & Pending PD/PI. The report now names the grant's lead: the contact PI, else the PI/MPI on the team, else the Lab PI when "My role" is PI or MPI. Before, it listed whoever the report was run for, so a postdoc's report named the postdoc as PD/PI.
• Reporting schedule for ongoing awards. The award form now pre-ticks the reporting schedule for any active award that doesn't have one, so an ongoing grant entered as Awarded gets its RPPR and close-out reminders. Closed awards still aren't offered it.
• Notice date. A new award's notice date is its decision date when one is recorded. An award entered late through the status dialog now lands in the right year in Funding awarded.
• Funding totals. Funding awarded and Total awarded count only applications marked Awarded, matching the success rate. Money no longer shows up for an application moved to Not funded that still has an award record.

Crashes and validation

• Leap day. The dashboard no longer fails on Feb 29; the 5-year window now uses relativedelta.
• Tasks on closed applications. Tasks on Withdrawn or Not pursued applications can be edited again. Those applications still aren't offered when creating a new task.
• CSV import. Problems are now reported row by row instead of failing the whole import with a server error. That covers values beyond the header, numbers too large for their column, NaN or Infinity, and probabilities outside 0–100. Extra decimal places are still accepted and rounded, as before.
• Long tags. Tags and critique themes longer than 60 characters are looked up by their stored (truncated) name, so reusing one no longer hits the unique constraint.
• Redirect. After claiming the Lab PI role, the redirect only follows a same-site next URL, via safe_next.

Testing

• New tests. 19 new tests across test_analytics, test_import (new), test_people and test_views. 15 of them fail on the previous code, covering all ten findings; the other 4 guard behavior that must not change.
• Full suite. 136 tests pass on PostgreSQL 16. check --deploy and makemigrations --check are clean.

🤖 Generated with Claude Code

https://claude.ai/code/session_01SAPwFhRyU83ey74tK1cYbJ

---

Generated by Claude Code