GitLab

added detailed comment on why that test cannot have nonempty pattern (we'd need to implement new parameter or change cutoff to 1 month from head block instead of from now())

[ABW]: removed duplicate tests that were moved to separate folder; updated to latest comment version

one test for get_content_replies moved to its own folder
updated to latest pattern versions

[ABW]: removed duplicate tests that were moved to separate folder; updated with latest pattern versions

[ABW]: changed tests and patterns to cover reblogging that actually happened (as test) before 5mln blocks

above tests also separated into single test .yaml files
similar thing done in other APIs just to have folders and examples ready

[ABW]: Patterns corrected manually where order depended on pending payouts (they would never match original especially that order within the same payout used to be random)
Usual differences accepted in reordered patterns: future cashout time, pending payouts, post_id
Some more tests marked as passing

Changed how error cases are handled (when error is unexpected or expected error is not there whole response is dumped) - some required regeneration of original patterns
Some patterns for tests that still don't work regenerated to fix key order
Updated comments for problematic tests (research results)
Some new bridge_api tests marked as passing
Order partially fixed in patterns for tests where that order depends on pending payout

accepted usual differences in patterns: rebranding, pending payouts, post_id, escaping in json_metadata, reputation, future cashout_time, order when depending on pending payout
updated comments on condenser_api new tests and marked as passing/failing properly

[ABW]: Added .orig.json for all tests (even if not different from pattern) to make it possible to compare with original response of old HiveMind/fat node
Accepted differences in patterns: future cashout_time, max_cashout_time (no longer relevant), rshares/weight related after first payout but before final, last_payout in the same case when payout was ineffective (that difference will last even at full sync), numerical instead of strings for long numbers (consistency)
Changed test comments to reflect current differences (mostly active, some payout related, still plenty of differences in order)

- jsondiff has a bug/feature: if you compare not empty json object with
  empty json object it will show no difference. Comparing empty with not
  empty will show difference
- jsondiff its not actively maintained (last commit 11 months ago)

Tests that only differed by last_payout now marked as passing
Creation time of accounts changed one block earlier to match what old FatNode was giving (unknown discrepancy between FatNode and old HiveMind)

slight differences in pending payout (expected) accepted in some patterns that had no other problems

- rebranding
- accepted different future cashout_time (prior to HF17 it was taken from root and we don't want to introduce hardfork differences to HiveMind)
- accepted different post_id
- accepted lack of double-escaped slashes in json_metadata (now is better)
- corrected indent for tavern's 2 spaces
- corrected cases of CLRF
Comparison script now removes .diff.json and .out.json files from previous run
Comparison script now generates .out.json with error message in case of error/no response (all failing tests have .out.json file)
Added comments to failing tests why they are failing
Added comments to tests with empty results why there is no other option

- Added xfail mark to failing tests

- Defined marks in proper way in pytest.ini.
- Added condenser_api pattern test yaml definitions.
- Changed xfail marks to failing marks.
- Update readme file with comments on marking.

- Added pattern validator
- Added patterns for 5000000 blocks
- Added bridge_api pattern tests - failing tests are marked with XFAIL
- Added marks to all tests