Let the Copernicus client say why a download failed

[chross22]

The symptom

A 52-day fetch failed on 11 days with an error that had the cause removed from it:

Error: 11 of 52 day(s) could not be downloaded:
  2026-07-15: Copernicus download failed after 3 attempt(s).
  ...
Client output:

Client output: followed by nothing.

Two causes, one mine

The dates were outside the dataset. cmems_mod_glo_phy_my_0.083deg_P1D-m covers 1993-01-01 → 2026-06-23. Every failing date was past that: a reanalysis runs months behind the present.

The reason was suppressed. Downloads ran at --log-level QUIET, which silences the client's explanation along with its chatter. The client says this at ERROR, and now does:

Some of your subset selection [2026-07-15, 2026-07-15] for the time dimension
exceed the dataset coordinates [1993-01-01, 2026-06-23]

Normal runs stay quiet — ERROR suppresses the INFO lines that made QUIET attractive.

Plus an explanation, not just a report

Out-of-range dates are the common cause, will fail on every retry, and the remedy is not obvious from a date range. When the client reports that, the error now adds:

Some dates are outside what this dataset covers. A reanalysis (_my_ in the dataset id) runs behind the present by months, so recent dates are not in it yet. Use mode = "forecast" for dates near or slightly ahead of today; note that the forecast reaches about ten days ahead, so dates beyond that are in no product.

Verification

Reproduced against the live API before and after — the same fetch now names the cutoff and the remedy.

• 2 regression tests: the default log level is ERROR and never QUIET; the hint fires when the client reports exceeded coordinates.
• R CMD check — Status: OK. Full suite green. Installed into both R 4.6.1 and 4.3.2.

🤖 Generated with Claude Code