Summary
Every config and auth flag on gateway source update is dead — the API rejects the request with config is not allowed. Source auth flags on create are rejected the same way for type WEBHOOK.
Reproduce
hookdeck gateway source update <id> --allowed-http-methods "GET,POST"
→ failed to update source: error: config is not allowed
hookdeck gateway source update <id> --config '{"allowed_http_methods":["GET","POST"]}'
→ failed to update source: error: config is not allowed
On create, for type WEBHOOK:
--api-key ... / --basic-auth-user ... / --hmac-secret ...
→ config.auth is not allowed
--webhook-secret does work on provider types (verified: STRIPE stores auth.webhook_secret_key).
The asymmetry
gateway destination update --config works fine — round-tripped path_forwarding_disabled and http_method successfully. So this is specific to sources.
Scope
Pre-existing, not a regression. Identical on fbb924a and the merged branch (the baseline dumps a raw 422; the merged build reports it more readably, which is an improvement but not a fix).
Open question
It is not clear from the CLI side whether this is:
- the CLI sending a shape the API no longer accepts,
- flags that should never have been offered for these source types, or
- an API-side restriction that should be lifted.
Worth confirming against the API before deciding whether the fix is to remove the flags, change what is sent, or change the API. If the flags cannot work, offering them is the bug.
Related
Filed by Claude on Phil's behalf, from release-candidate regression testing.
Summary
Every config and auth flag on
gateway source updateis dead — the API rejects the request withconfig is not allowed. Source auth flags oncreateare rejected the same way for type WEBHOOK.Reproduce
On
create, for type WEBHOOK:--webhook-secretdoes work on provider types (verified: STRIPE storesauth.webhook_secret_key).The asymmetry
gateway destination update --configworks fine — round-trippedpath_forwarding_disabledandhttp_methodsuccessfully. So this is specific to sources.Scope
Pre-existing, not a regression. Identical on
fbb924aand the merged branch (the baseline dumps a raw 422; the merged build reports it more readably, which is an improvement but not a fix).Open question
It is not clear from the CLI side whether this is:
Worth confirming against the API before deciding whether the fix is to remove the flags, change what is sent, or change the API. If the flags cannot work, offering them is the bug.
Related
--configmeans two different things across gateway and outpostFiled by Claude on Phil's behalf, from release-candidate regression testing.