support LegacyTests in testeth

[winsvega]

update the test subrepo to latest develop
support testeth to run LegacyTests (all tests for forks <= Constantinople. inc. GeneralStateTests and its blockchain form at the moment)

./testeth -t LegacyTests -- --all

note this tests only run on linux-gcc6 yaml because --all option is enabled

[gumb0]

Some tests are failing

[codecov-io]

Codecov Report

Merging #5735 into master will increase coverage by 0.31%.
The diff coverage is 95.3%.

@@            Coverage Diff             @@
##           master    #5735      +/-   ##
==========================================
+ Coverage   63.35%   63.67%   +0.31%     
==========================================
  Files         353      356       +3     
  Lines       30352    30469     +117     
  Branches     3395     3390       -5     
==========================================
+ Hits        19230    19401     +171     
+ Misses       9886     9843      -43     
+ Partials     1236     1225      -11

[gumb0]

this is just a rant, we won't change it now, but this should have been called from the test body instead of weirdly from fixture constructor; for that you could have created a macro wrapping BOOST_AUTO_TEST_CASE macro always calling it from the body
(maybe do that in retesteth if it's any similar there)

[winsvega]

Please explain better the idea.

[gumb0]

Don't run the tests from the constructor, run them from the test body.

#define RETESTETH_AUTO_TEST_CASE(casename) BOOST_AUTO_TEST_CASE(casename) { m_suite.runAllTestsInFolder(casename); }

RETESTETH_AUTO_TEST_CASE(stCallCodes) 

...anyway it's better to get rid of boost.test altogether

[winsvega]

I think it will get more confusing if to continue refactoring it that way

[gumb0]

I don't see legacy tests running on AppVeyor, do they run? (maybe just the suite name is not logged?)

[gumb0]

Add item to changelog

[winsvega]

In AppVeyor it just runs testeth that output just the casenames without the tree structure.
The tests get executed. (no --all flag is set )

[gumb0]

Add item to changelog

[gumb0]

more comments

[gumb0]

looks better now, a couple of small things left

[gumb0]

LegacyTests/Constantinople/BCGeneralStateTests are reported as passed by ctest, but they are actually skipped, aren't they?
But it was similar before for blockchain tests generated from state tests, right?

[winsvega]

LegacyTests/Constantinople/BCGeneralStateTests get executed if ctest run testeth with --all flag. otherwise it is skipped same as stTimeConsuming tests.

I've checked the scripts. I think it was macOS that runs all tests with --all flag.

[gumb0]

Please squash at least first 4 commits, to make the anonymous ones disappear

[winsvega]

is there a command to squash commits in the history not touching the recent ones?

[gumb0]

depends on what you mean by not touching
git rebase -i HEAD~9 and set the first ones to squash and the last ones to pick
(all commits will be rewritten anyway, resulting in new commit hashes)

[winsvega]

looks like I can't sqush the first unnamed commit. you could only squash following commits into previous.

check it out on branch: legacytests_squash
if its ok. I will push it to this branch.

[gumb0]

That branch is ok, it shows first commit to be authored by Wins and committed by @winsvega

Now GCC compiler is crashing, could you also cherry-pick this commit here 874a5e7

[winsvega]

have you just squashed the commits in this branch ok?

[gumb0]

Yes, it's fine here now

[gumb0]

@chfast Is it ok to merge this with BUILD_PARALLEL_JOBS: 2 for GCC?

[chfast]

@chfast Is it ok to merge this with BUILD_PARALLEL_JOBS: 2 for GCC?

Yes