Upload test results to the database #4897
Conversation
|
@fabriziomello, @pmwkaa: please review this pull request.
|
Oh wow, what is this stuff 😮 Please disregard it. |
Codecov Report
@@ Coverage Diff @@
## main #4897 +/- ##
==========================================
- Coverage 90.90% 89.52% -1.38%
==========================================
Files 224 226 +2
Lines 42806 50999 +8193
==========================================
+ Hits 38912 45659 +6747
- Misses 3894 5340 +1446
Continue to review full report at Codecov.
|
| - name: Get date for build caching | ||
| id: get-date | ||
| run: | | ||
| echo "date=$(date +"%m")" >> $GITHUB_OUTPUT |
There was a problem hiding this comment.
hmm i think weekly or even daily would be better
| test_duration float | ||
| ); | ||
|
|
||
| create unique index on test(job_date, test_name); |
There was a problem hiding this comment.
I think it's convenient to have unique names for our tests. We had some duplicates (e.g. an isolation and a regression test with the same name), so I renamed them.
| print buf[(NR + i - to_print) % max_context_lines] | ||
| } | ||
|
|
||
| printf("c %04d: %s\n", NR, $0); |
There was a problem hiding this comment.
The c %04d: part is for debugging. I'd like to leave it in for a while, until I'm absolutely sure that this diff snipping code works. It doesn't harm readability too much, and actually helps notice errors, because it has line numbers.
|
Hmm not a fan of this being in the matrix builder, can we split this into separate file to keep the matrix builder cleaner |
You mean, the flaky check? I don't know, logically it has to be somewhere around the matrix builder because it adds a special configuration and computes the installcheck options. I can split out the calculation of the changed tests into a separate procedure. Or even into a separate file, although to my taste that's an overkill. Sorry for this PR being confusing, I already do three things here -- fixing flaky check, fixing the LLVM options like you asked me before, and finally adding the upload itself. Maybe it's better to split it up. |
Four things, also changing how the Postgres builds are cached. |
|
Factored some stuff out: |
|
|
||
| with open(os.environ['GITHUB_OUTPUT'], 'a') as output: | ||
| print('regression_diff=true', file=output) | ||
| find . -name regression.diffs -exec cat {} + > regression.log |
There was a problem hiding this comment.
Why remove the python code here? for convenience or is there another reason?
There was a problem hiding this comment.
I wanted to make it as similar as possible to the other checks like Linux. This way it's easier to maintain, I can just diff those files and copy the fragments of the code, no need to write same thing anew in another language.
akuzm
force-pushed
the
aku/ci-db
branch
8 times, most recently
from
b892f47
to
04a08cf
Compare
This will help us find the flaky tests or the rare failures.
This will help determine the most flaky tests.