Fix ECDSA with TLS 1.0 / TLS 1.1 #187
Conversation
Handshake failed with "bad SignatureRSA for ecdhparams" when the signature is actually SignatureECDSA. The default hash algorithm for ECDSA should be SHA-1.
It is usually installed as a dependency of the test suite, so when RUNTEST=0 the package was missing.
|
The job failures could be the result of https://www.traviscistatus.com/incidents/sxrh0l46czqn |
Related to Travis job #233.4, which picked unsupported version async-1.4 for a reason I don't understand yet.
|
Everything is back to normal and the build now reports why package async causes a failure with GHCVER=head (base-4.10 not allowed yet). For reference the issue was a client error "bad SignatureRSA for ecdhparams" So I'll merge this in a couple of days unless someone objects. |
|
Sorry for the delay. |
|
Great, you'll see the changes are small. I just noticed the additional commit in #185, so please tell me if you are OK to merge your branch network-version too. |
|
I understand that this is just a copy-and-paste mistake. |
|
I sent a pull request for |
|
Everything is fine, thank you. |
I see two explanations for TestClient failures in Travis builds:
the default hash algorithm for ECDSA should be SHA-1 (just like for DSA)
as described in RFC 4492 § 5.4
what looks like a copy/paste mistake in the definition of two ECDSA ciphers