tests: NAF and splitK in crypto.secp256k1.curve #74
Conversation
|
Bah, dataclasses were introduced in Python 3.7 . Shall I use a standard class or a dictionary? It's just a few lines in a test. EDIT: another possibility is to add the EDIT2: alas, Poetry won't install I'll use a dictionary. |
| @@ -210,10 +210,10 @@ def splitK(self, k): | |||
| # being a bottleneck. | |||
| # c1 = round(b2 * k / n) from step 4. | |||
| # Rounding isn't really necessary and costs too much, hence skipped | |||
| c1 = (self.b2 + k) // self.N | |||
| c1 = (self.b2 * k) // self.N | |||
There was a problem hiding this comment.
This (and the similar one below) is a bugfix: the original Go code has a bigInt.Mul here, and tests fail without this change.
There was a problem hiding this comment.
Weird that we never hit an issue with this with signature verification. Hmmm.
I won't claim I understand it either. I did have a quick look at the algorithms in the book mentioned in the Go code, but most of the work is translating the Go code and tests anyway. |
| @@ -210,10 +210,10 @@ def splitK(self, k): | |||
| # being a bottleneck. | |||
| # c1 = round(b2 * k / n) from step 4. | |||
| # Rounding isn't really necessary and costs too much, hence skipped | |||
| c1 = (self.b2 + k) // self.N | |||
| c1 = (self.b2 * k) // self.N | |||
There was a problem hiding this comment.
Weird that we never hit an issue with this with signature verification. Hmmm.
decred/decred/util/helpers.py
Outdated
| return 0 | ||
|
|
||
|
|
||
| random.seed(0) |
There was a problem hiding this comment.
Can't set the random seed in production code. I'm hesitant to add any functions here that are just for tests. How about a test utilities module somewhere?
There was a problem hiding this comment.
Can't set the random seed in production code. I'm hesitant to add any functions here that are just for tests.
Makes sense.
How about a test utilities module somewhere?
Test code does not live in packages and modules. Test discovery is performed by pytest according to its own rules. For cases like this, pytest offers fixtures, I'll change randBytes and sign as such.
The remaining question is whether setting the seed should be done in the fixture or in the test that uses it: I lean towards the latter.
- test helpers "randBytes" and "sign" removed from util/helpers.py and moved to a conftest.py file in the tests/ root directory, as pytest fixtures; - test_curve.py changed to use the new fixtures; - unit/util/test_database.py restored to its original state.
Part of #70.
Converted the
crypto.secp256k1.curveunit tests to pytest format.Added tests for
NAFandsplitKincrypto.secp256k1.curve, and fixed a double bug insplitK.Moved
randBytestoutil.helpersand added asignfunction (that Python unfortunately lacks).Changed
curve.lamdatocurve.lambda_(trailing underscores are the usual Python convention for avoiding name conflicts).