Fix issue #211 #214
Fix issue #211 #214
Conversation
…test/bats subrepo: subdir: "test/bats" merged: "a7106392" upstream: origin: "https://github.com/bats-core/bats-core.git" branch: "master" commit: "a7106392" git-subrepo: version: "0.4.5" origin: "???" commit: "???"
- Invert default of when to run tests. Default is that they are OFF. - Simplify test framework to remove "variant" tests from scope. - Simplify by no longer bake "fast" vs "slow" or "unit" vs "integ" into test source file name. - Bring `waf_unit_test` into `waft/wcb_unit_test` and move test argument handling there. - Test related arguments will set during `configure` are saved to `build`. - Add new feature to measure, store and report the run time for each test. - Add new feature to exclude re-running slow tests based on a maximum duration limit. - Update `framework.org` to match.
- Make the repo "live" only and entirely under build/tests and remove complexity of searching a "test data path". - Add fillrepo and packrepo "wcb" commands to automate some management of test data repository files. - Remove wct-bats.sh functions related to the old way of locting test data files, add/improve ones for the new way. - Further purge the concept of "variant" tests and likewise downplay their opposites of "atomic" tests. - Sync documentation and tests with these code/concept shifts.
Downgrade issue202 bats to a check as it is too long running
The "measure" is found by applying the connected components algorithm to a special per-slice, per plane graph holding nodes representing blobs and channels. The nodes of this graph hold the vertex descriptor of the original cluster graph node. The bug was due to a channel being added to this special graph each time it was seen across the blobs. Thus, by construction, the connected components (ie, measures) always had exactly one blob. The fix was to only add any given channel exactly once. The following test is extended to check this bug: ``` $ bats img/test/test-wct-uboone-img.bats test-wct-uboone-img.bats ✓ create graph ✓ check log files ✓ inspect blobs ✓ dump blobs ✓ at least one multi-blob measure 5 tests, 0 failures ```
|
@brettviren, I got some error trying to run the uboone-img test: Which is related to no trace summary. Maybe you forgot to update the frame input here?
|
|
The sim/sigproc tests seem OK. Just the sigproc results fluctuate again.
More notes: pr214.pdf The Charge Solving results have much better consistency with WCP for multi-blob solving now: |
|
Hi @HaiwangYu
I think you are right. At least, the I've just refreshed the Hopefully your running of the test will work now. To assure fresh download, you may need to: Aside comment: This out-of-sync problem would have been avoided if I remembered to do a |
|
@brettviren, I still get an error which seems to be related to the numpy json diff test. But I think we could merge this PR first? |
|
Yes, I think it is okay to merge. I'll just go ahead and do that. I'll look at checking the json vs numpy format test more. |
This needs to be applied on top of PR #199 (aka, it includes all those commits).
See last commit for details of the fix.