better test coverage for internal modules #146

tomato42 · 2019-10-27T01:00:31Z

preparation for #127
make the test coverage more readable and make it use pytest or unittest asserts for checks
where possible, also use pytest parametrisation or hypothesis

coveralls · 2019-10-27T01:16:26Z

Coverage increased (+1.9%) to 94.858% when pulling 89fe31d on tomato42:ecdsa-test-cov into fa7655c on warner:master.

don't use six.print_ - not needed on py2.6 sort imports in conventional way

make the test with different keys, messages and nonces reproducible use all the curves/generators defined in the test (not only P-192)

use pytest.parameterize to execute the Known-Answer Tests for the NIST P-256 curve use a bit more sane formatting for the messages in the tests

use hypothesis (and larger numbers) for testing the inverse_mod function

make the jacobi test into a hypothesis test increase the range tested add test to verify factorization, since it's used in the jacobi test

extend the testing for large primes with hypothesis use pytest.parametrize to test with small primes

move sanity checks for gcd() to unittest add hypothesis tests for gcd

the function is just a wrapper around pow() builtin, so don't use it, mark it deprecated, exclude from coverage (as we're not testing it)

add hypothesis test to lcm

simo5

Some of the tests sound a bit redundant, when all you are doing is testing if the math library works correctly, but besides that I see no issues with this PR.

simo5 · 2019-10-31T13:22:42Z

src/ecdsa/test_numbertheory.py

+                         "Hypothesis 2.0.0 can't be made tolerant of hard to "
+                         "meet requirements (like `is_prime()`)")


Not really sure I understand what this means, might be clarified, for people not too familiar with hypothesis

That is, what is the actual issue ?

when you select random integers, few of them will be primes, so the test case generation takes a lot of time; hypothesis doesn't like it when it takes few seconds
with new versions, we can disable those checks, but with old (the oldest supported on py2.6) we can't

Pleas mention that the issue is a time issue then.

tomato42 · 2019-10-31T13:35:10Z

when all you are doing is testing if the math library works correctly

yes, the maths library that is part of this package... there is no lcm(), gcd() or inverse_mod() in standard python... or am I missing something?

simo5 · 2019-10-31T14:07:00Z

when all you are doing is testing if the math library works correctly

yes, the maths library that is part of this package... there is no lcm(), gcd() or inverse_mod() in standard python... or am I missing something?

There is math.gcd() but lcm() is indeed not there, although lcm() is so trivial to create once you have gcd(). I thought lcm() was also there but now that I look at it, it turns out I also implemented it recently for some test elsewhere :-D
So my comment I guess apply only to gcd() at this point.
It's not a big deal.

tomato42 · 2019-10-31T14:22:03Z

There is math.gcd()

Indeed, but not in older versions:
https://docs.python.org/3/library/math.html#math.gcd:

New in version 3.5.

probably we can use it to speed up newer versions, but the old code needs to remain

ok, thanks for the review!

tomato42 added the maintenance issues related to making the project usable or testable label Oct 27, 2019

tomato42 self-assigned this Oct 27, 2019

tomato42 force-pushed the ecdsa-test-cov branch from 226e8d0 to 2a8195a Compare October 27, 2019 23:48

tomato42 changed the title ~~better test coverage for ecdsa module~~ better test coverage for internal modules Oct 27, 2019

tomato42 force-pushed the ecdsa-test-cov branch 5 times, most recently from abc8202 to f4da621 Compare October 28, 2019 11:51

tomato42 mentioned this pull request Oct 28, 2019

Minor speed-up #147

Merged

tomato42 added this to the v0.14 milestone Oct 28, 2019

tomato42 force-pushed the ecdsa-test-cov branch 5 times, most recently from 5ac843b to 6f712e1 Compare October 30, 2019 13:28

tomato42 marked this pull request as ready for review October 30, 2019 13:29

tomato42 added 12 commits October 30, 2019 20:45

test_ecdsa: cleanup imports

bed03a2

don't use six.print_ - not needed on py2.6 sort imports in conventional way

test_ecdsa: use hypothesis for random sign and verify

95fb82f

make the test with different keys, messages and nonces reproducible use all the curves/generators defined in the test (not only P-192)

test_ecdsa: parametrise KATs for P-192

1fc153d

use pytest.parameterize to execute the Known-Answer Tests for the NIST P-256 curve use a bit more sane formatting for the messages in the tests

test_ecdsa: parametrize point validity test

da6b9f6

test_ecdsa: clean up the tests from X9.62

e36576b

test_numbertheory: hypothesis for inverse_mod

b17549e

use hypothesis (and larger numbers) for testing the inverse_mod function

test_numbertheory: test jacobi and factorization functions

7b394e8

make the jacobi test into a hypothesis test increase the range tested add test to verify factorization, since it's used in the jacobi test

test_numbertheory: test square_root_mod_prime

9e09a8a

extend the testing for large primes with hypothesis use pytest.parametrize to test with small primes

test_numbertheory: test next_prime

fa9dded

test_numbertheory: test gcd

518cfae

move sanity checks for gcd() to unittest add hypothesis tests for gcd

numbertheory: deprecate modular_exp

1472274

the function is just a wrapper around pow() builtin, so don't use it, mark it deprecated, exclude from coverage (as we're not testing it)

test_numbertheory: test lcm

dc381a6

add hypothesis test to lcm

tomato42 force-pushed the ecdsa-test-cov branch from 6f712e1 to dc381a6 Compare October 30, 2019 19:46

simo5 approved these changes Oct 31, 2019

View reviewed changes

test_numbertheory: slightly nicer comments

89fe31d

tomato42 merged commit 8a4a1f9 into tlsfuzzer:master Oct 31, 2019

tomato42 deleted the ecdsa-test-cov branch October 31, 2019 15:02

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

better test coverage for internal modules #146

better test coverage for internal modules #146

tomato42 commented Oct 27, 2019 •

edited

Loading

coveralls commented Oct 27, 2019 •

edited

Loading

simo5 left a comment

simo5 Oct 31, 2019

simo5 Oct 31, 2019

tomato42 Oct 31, 2019 •

edited

Loading

simo5 Oct 31, 2019

tomato42 commented Oct 31, 2019

simo5 commented Oct 31, 2019

tomato42 commented Oct 31, 2019

		"Hypothesis 2.0.0 can't be made tolerant of hard to "
		"meet requirements (like `is_prime()`)")

better test coverage for internal modules #146

better test coverage for internal modules #146

Conversation

tomato42 commented Oct 27, 2019 • edited Loading

coveralls commented Oct 27, 2019 • edited Loading

simo5 left a comment

Choose a reason for hiding this comment

simo5 Oct 31, 2019

Choose a reason for hiding this comment

simo5 Oct 31, 2019

Choose a reason for hiding this comment

tomato42 Oct 31, 2019 • edited Loading

Choose a reason for hiding this comment

simo5 Oct 31, 2019

Choose a reason for hiding this comment

tomato42 commented Oct 31, 2019

simo5 commented Oct 31, 2019

tomato42 commented Oct 31, 2019

tomato42 commented Oct 27, 2019 •

edited

Loading

coveralls commented Oct 27, 2019 •

edited

Loading

tomato42 Oct 31, 2019 •

edited

Loading