← all records

Driver config and CLI cleanup

pitch-4f29ab · in warm_bubble

  • No No-gos section. The pitch template asks for one — it is what keeps the appetite honest.

Problem

The configuration system landed in cycle 37 (pitch-370009, #1391). Using it turned up three things that were wrong with the driver around it, which OngChia filed on 24.08.2026 as issue-79be09, issue-28b2f7 and issue-f89c58:

  • do_prep_adv was a DriverConfig parameter although it is not a choice: it is required whenever tracer advection is on, so the user had to keep two settings consistent by hand.
  • Output came out one time step early and never included the initial condition, because the monitor asked whether the next step was an output step rather than the current one.
  • The CLI could read only ICON-generated JSON namelists, had no tests, and exposed enable_output and output_mode as options that could silently override the YAML driver configuration.

None of this reached the cycle-38 betting table, where OngChia appears only as support on other people's rows. It is nonetheless what he worked on during the cycle, so it is written here as a bet of its own.

Appetite

Two weeks. Three bounded fixes against a config system that is already built.

Solution

One PR per issue, each described in the issue it answers.

Rabbit holes

The CLI rework (#1466) is the only one of the three that is not a contained fix: it removes the legacy JSON path and moves the f90-to-yaml conversion out to scripts, so it changes how everyone invokes the driver.

No-gos

Progress

  • #1465 remove do_prep_adv from config -- merged 15.09.2026, answers issue-79be09
  • #1467 fix output time step -- merged 17.09.2026, answers issue-28b2f7
  • #1466 driver CLI reads config from yaml, legacy jsons removed, f90->yaml conversion moved to scripts -- open, no reviewer yet, answers issue-f89c58