R4R: Add proof verification

[abelliumnt]

The merkle proof can be divided into two category: existence proof and absence proof. They have difference verification methods. Current, the proof verification will use wrong method to verify absence proof if the query key doesn't exist.

[zmanian]

Can we write a test case for this?

[abelliumnt]

I added a new commit to enable distrust mode in lcd test. Then the currently lcd test can cover the code for proof verification.

[rigelrozanski]

looks like lcd test is failing and should be fixed (in test_cover)

[cwgoes]

This is probably failing because of the upstream issue - tendermint/tendermint#2862.

[mossid]

The test is failing because the validator set is different in some of the tests. Before TestBonding, InitializaTestLCD takes 1, in result generating same validator set. In TestBonding, the argument becomes 2 to test validator bonding. tests.WaitForHeight queries http://localhost:port/blocks/latest to the lcd. The problem is that the lcd is not reset at the end of each TestXXX, so the previous tests' trusted commits are not cleaned up. So when the validator set is different from the previous ones(TestBond in this case), the lcd query is requested to proof a signed header whose validator set is different from the stored one with the same height, returns error, and panics on tests/util.go when attempt to unmarshal it as a json value.

[jackzampolin]

An alternate implementation: #3236

[jackzampolin]

@mossid how do we fix that issue?

[abelliumnt]

@mossid
In lcd tests, the lcd home directories are readomly generated. So it is impossible that two lcd processes share the same home directory.
Besides, the failed unit test is unbondingTest, if you launch the single test, it will pass. If you run all tests together, then it will fail. That's very strange.

[jackzampolin]

@HaoyangLiu The tests are dependent on eachother. I did some work trying to parallelize them about a month ago without much luck.

[jackzampolin]

@HaoyangLiu Is this PR ready to merge? It appears to be still failing tests.

[abelliumnt]

@jackzampolin
I still working with @mossid to fix the failed unit test.

[jackzampolin]

@HaoyangLiu Thanks for the update!

[cwgoes]

OK, I tried to make a downstream PR - https://github.com/HaoyangLiu/cosmos-sdk/pull/15 - but it looks like CI isn't working, so I'm just going to split this into a separate branch.

[cwgoes]

Closing in favor of #3449 - but thank you @HaoyangLiu!

[abelliumnt]

@cwgoes
Great! Recently I was busy with irishub development, so I missed to fix the failed lcd tests. Now Are all lcd tests passed?

[cwgoes]

@cwgoes
Great! Recently I was busy with irishub development, so I missed to fix the failed lcd tests. Now Are all lcd tests passed?

Yes - fixed in #3449, now merged.