release v0.7.1
#1176
release v0.7.1
#1176
Conversation
Fix #1146 block state root issue
Bls G2 padding issue
Fuzzing utilities package / SSZ decoding for spec
generate coverage reports in make test, open as html site
There was a problem hiding this comment.
The block state-roots are fixed now, but for review I went through all of it again, and think it's better to not rely on the zero-values for 1 special case: when we overwrite the latest-header, we also implicitly change the state-root to zero. And then there is an undocumented dependency where the next process_slot has to be the very first thing to touch the state, as it hashes it to produce the state-root to put into latest-header. Also, there is a zero-check, but it's not incorrect to not overwrite when it's non-zero. Instead, it should be equal. All in all correct, and we can just merge v0.7.1, but it could be cleaner. What do you think @JustinDrake, would you like to pick this up?
| @@ -23,3 +26,13 @@ def get_state_root(spec, state, slot) -> bytes: | |||
| """ | |||
| assert slot < state.slot <= slot + spec.SLOTS_PER_HISTORICAL_ROOT | |||
| return state.latest_state_roots[slot % spec.SLOTS_PER_HISTORICAL_ROOT] | |||
|
|
|||
|
|
|||
| def state_transition_and_sign_block(spec, state, block): | |||
There was a problem hiding this comment.
Verified we are not outputting any blocks in the tests not produced by this function. appyl_empty_block is not used for output anywhere. But we may want to remove its return in the future, so we don't accidentally get a block with faulty state root in there.
| @@ -2,3 +2,4 @@ | |||
| pytest>=3.6,<3.7 | |||
| ../config_helpers | |||
| flake8==3.7.7 | |||
| pytest-cov | |||
There was a problem hiding this comment.
yay for coverage reports. Don't forget to make install_test to enable it. Good for changelog.
A few critical fixes:
state_rootin blocks in tests (Invalid sanity test block state root #1146)Additional changes: