Every PR runs all 190 specs — is there a --since for sdods run?

Asked 21 May 20263,280 views2 answers

19votes

PR checks take 40 minutes because we run the whole suite on every push, and a typical PR touches one module. I was expecting something like

bash

sdods run -p portal --since origin/main

but that is not a flag. There is clearly change-impact analysis somewhere — is it reachable from run?

asked by Thom Vasseur5,470 · 21 May 2026

2 answers

18votes
It is not. sdods run has no --since and no equivalent. The analysis exists, but only as the analyze_change_impact tool on the MCP server — and nothing feeds its output back into the runner. Open issue; the ask is exactly your flag.
Two-step version that works today: call the tool from CI, take the modules or tags it reports, and pass them to run:
bash
```
sdods run -p portal -m <module> -m <module> -t @regression
```
That does cut the time. It just makes your CI job responsible for wiring together two things that should be one.
answered by Priya Venkateshmaintainer29,604 · 21 May 2026
11votes
We do that, and the awkward part is not the plumbing, it is the fallback.
When the tool returns nothing — a change it cannot map, a config-only PR, a failure in the step itself — you have to choose between running everything (and losing the saving on exactly the changes you understand least) and running nothing. Get that branch wrong and you have a green PR check that tested nothing, quietly, for weeks. Ours defaults to the full suite on an empty result and shouts about it in the job summary.
If you build this, spend your care there rather than on the selection.
answered by Marijn Kuiper18,840 · 22 May 2026

Have a different question? Ask it.