send: return error for v1 URIs in v2 sender

[vbrekher]

SenderBuilder::new in the v2 sender currently panics when it receives a v1 Payjoin URI.

Make the constructor fallible instead and return a normal build error for unsupported v1 URIs. The FFI constructor propagates the same failure through SenderInputError::Build.

This keeps the v1 sender behavior unchanged while making v2 safe to use with untrusted Payjoin URIs.

AI assistance

ChatGPT was used to inspect the codebase, develop the patch, add tests, run validation, and draft this PR description.

Closes #1849

[coveralls]

Coverage Report for CI Build 34412200999

Coverage increased (+0.02%) to 86.663%

Details

• Coverage increased (+0.02%) from the base build.
• Patch coverage: 1 uncovered change across 1 file (24 of 25 lines covered, 96.0%).
• No coverage regressions found.

Uncovered Changes

File	Changed	Covered	%
payjoin/src/core/send/error.rs	2	1	50.0%
Total (3 files)	25	24	96.0%

Coverage Regressions

No coverage regressions found.

---

Coverage Stats

Relevant Lines:	16571
Covered Lines:	14361
Line Coverage:	86.66%
Coverage Strength:	341.41 hits per line

---

💛 - Coveralls

[chavic]

Hi @vbrekher , thanks for picking this up. The added type UnsupportedVersion probably needs to match the same gating as the arm that uses it. Without that properly gated, I expect clipy to fail when only v2 or v1 is enabled

[vbrekher]

Good catch, thanks. I gated UnsupportedVersion and the corresponding Display/Error::source match arms with #[cfg(feature = "v1")] so they match the constructor arm that uses the variant. Pushed the fix; the new Actions runs are currently waiting for approval.

[chavic]

@vbrekher, logs show another unrelated error; you'll need to update the JS assertion too

[vbrekher]

Thanks, found it in the JS integration log. The v1 URI now fails at SenderBuilder construction with the typed SenderInputError.Build, so the old /RuntimeError/ assertion was stale. I updated the assertion to expect SenderInputError.Build and removed the now-unreachable buildRecommended(...) call from that check. Pushed as d94e568.

[chavic]

Good catch, thanks. I gated UnsupportedVersion and the corresponding Display/Error::source match arms with #[cfg(feature = "v1")] so they match the constructor arm that uses the variant. Pushed the fix; the new Actions runs are currently waiting for approval.

And you'll actually want all(feature = "v1", feature = "v2"), since only having one should still fail. Can edit that commit rather than adding a new one. f85b369, which is the space removal, also belongs to an edit to a commit rather than being its own commit

[vbrekher]

Yep, fixed. The gate is now all(feature = "v1", feature = "v2"), and I cleaned up the branch history so the follow-up fixes are folded into the original commit. I also rebased it onto the current master, so the PR is back to one commit and no longer drifted. New Actions runs are waiting for approval.