Invalid target header name error no longer indicates which line was rejected #247

New Issue

clawbot · 2026-08-20T10:29:52+02:00

clawbot commented

2026-08-20 10:29:52 +02:00

Surfaced during review of #242. Not milestoned for 1.0; filed so the finding is recorded rather than lost in a review thread.

errHeaderNameInvalid now quotes nothing at all. That is a direct and correct consequence of #233, which required the error not to echo a possible secret — so this is a trade-off that landed, not a defect in that work.

The cost is operator UX: someone submitting a multi-line headers field gets no indication of WHICH line was rejected, and must bisect by hand.

Definition of done:

An operator submitting several header lines, one of which has an invalid name, can tell which line was rejected.
The error still never echoes a header VALUE, and never echoes raw bytes of the name — the property #233 established must not regress. A line number, an ordinal, or a strictly sanitised rendering of the name are all acceptable; the raw name is not.
A test asserts both halves: the rejected line is identifiable, and no secret material appears in the message.

Surfaced during review of https://git.eeqj.de/sneak/webhooker/pulls/242. Not milestoned for 1.0; filed so the finding is recorded rather than lost in a review thread. `errHeaderNameInvalid` now quotes nothing at all. That is a direct and correct consequence of https://git.eeqj.de/sneak/webhooker/issues/233, which required the error not to echo a possible secret — so this is a trade-off that landed, not a defect in that work. The cost is operator UX: someone submitting a multi-line headers field gets no indication of WHICH line was rejected, and must bisect by hand. Definition of done: - An operator submitting several header lines, one of which has an invalid name, can tell which line was rejected. - The error still never echoes a header VALUE, and never echoes raw bytes of the name — the property https://git.eeqj.de/sneak/webhooker/issues/233 established must not regress. A line number, an ordinal, or a strictly sanitised rendering of the name are all acceptable; the raw name is not. - A test asserts both halves: the rejected line is identifiable, and no secret material appears in the message.

~~clawbot referenced this issue 2026-08-20 11:07:50 +02:00~~

Milestone 1.0.0: internet-facing readiness #111

clawbot commented

2026-08-24 00:58:28 +02:00

Dismissed by execution during the deployability audit. The line number is present on both paths.

Multi-line header config with a bad line 3 returns exactly Invalid headers: line 3: header name must be a valid HTTP token, on create and on edit. Duplicate and reserved-header errors carry it too: Invalid headers: line 2: header given more than once: "X-Dup" and Invalid headers: line 2: header is set by the delivery engine and cannot be overridden: "Host".

Closing.

Dismissed by execution during the deployability audit. The line number is present on both paths. Multi-line header config with a bad line 3 returns exactly `Invalid headers: line 3: header name must be a valid HTTP token`, on create and on edit. Duplicate and reserved-header errors carry it too: `Invalid headers: line 2: header given more than once: "X-Dup"` and `Invalid headers: line 2: header is set by the delivery engine and cannot be overridden: "Host"`. Closing.

clawbot closed this issue

2026-08-24 00:58:35 +02:00

clawbot referenced this issue

2026-08-24 00:59:13 +02:00

1.0/mvp #33

clawbot referenced this issue

2026-08-24 03:34:45 +02:00

Record landed 1.0.0 work and drop the removed signing roadmap in TODO.md #284

Sign in to join this conversation.