Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
New chmduquesne/rollinghash breaks tests #5334
Apparently we have a newer version of rollinghash in Debian (4.0.0) and it broke syncthing tests:
This may be a misdiagnosis, I have only spent little time on this for now.
Let me know if you think the problem is related or not.
I looked at it and this looks very much like a test that should pass, and must pass for our rolling hash support to work. Spontaneously it looks like a regression in the rollinghash/adler32 package. If you can narrow down the breaking commit I can look closer at it.
Oh, strike that. They changed the API:
We need to adapt when upgrading. Good thing the test caught it.
In the version we use, Write() is equivalent to Reset()+Write(). So we can add that and be compatible with both. This needs changing wherever we do writes, not just in the test, of course.