Codecov Report

Merging #1051 (c4f13e0) into master (3274cba) will increase coverage by 0.11%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #1051      +/-   ##
==========================================
+ Coverage   93.97%   94.09%   +0.11%     
==========================================
  Files         179       86      -93     
  Lines       25140    12235   -12905     
  Branches     2613     1230    -1383     
==========================================
- Hits        23626    11512   -12114     
+ Misses       1210      582     -628     
+ Partials      304      141     -163     

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 3274cba...c4f13e0. Read the comment docs.

Thanks, I don't often get contributions to the C code! Sorry it's taken me so long to respond. Do you have a way to reproduce the problem this is fixing?

BTW, separately, I'd be interested to hear about the plugin you are working on.

I don't have a simple way of reproducing the problem. I discovered it when running liberapay's test suite with the experimental coverage plugins, and I confirmed that the changes in this branch fix the problem, but I didn't write test code that reproduces it and could be used to prevent regressions in the future.

The plugins I worked on are for Jinja2 and simplates. I obtained some encouraging results, but the work isn't finished, I paused after I opened this pull request.

I've rebased this branch on master, replacing frame->f_lineno with PyFrame_GetLineNumber(frame) like in 7e5e28f.

since this is so old, updates have happened since, and 3 checks are not being met ... is this something that can be closed?

This patch is still needed. I have now rebased it on master once again, and confirmed that it still fixes the problem.

The patch has become a bit smaller because its previous version modified the CTracer_check_missing_return function which was dropped in 8021196.

The patch has also been modified to use MyFrame_GetCode(frame) instead of frame->f_code, per ab48a7b.

Can you give me very specific instructions for cloning and running the test suite, and specific pointers to places where the coverage information is wrong? I took a quick look in the liberapay repo and couldn't find a coverage plugin.

The unfinished coverage plugins are now publicly available in the coverage-plugins branch of the liberapay.com repository.

To run the tests, a PostgreSQL server is required. By default the tests try to use the liberapay_tests database of the local server. Once the database is ready, running the tests with coverage tracking should be as simple as running make pytest-cov.

Results with the unpatched version of coverage

1. The pytest output contains quite a few warnings like this:

   simplate_coverage.py:241: UserWarning: got out of bounds line numbers: [136] > 112
   simplate_coverage.py:323: UserWarning: got out of bounds line numbers for file www/on/%platform/index.spt: [(-126, 87), (136, -62)] > 112
   

2. The htmlcov/index.html file reports a total of 3748 non-covered ("missing") lines.
3. The HTML coverage report for the templates/exceptions/EmailAddressError.html file erroneously claims that line 26 has been run while the surrounding lines in the same code branch haven't been.
   

Results with the patch

1. There are no out of bounds warnings in the pytest output.
2. The htmlcov/index.html file reports a total of 3857 non-covered ("missing") lines.
3. The HTML coverage report for the templates/exceptions/EmailAddressError.html file correctly shows that line 26 has not been run.