Обсуждение: differential code coverage
Hi, For quite a while I'd been wishing to see *differential* code coverage, to see what changed in code coverage between two major releases. Unfortunately lcov didn't provide that. A few months ago a PR for it has been merged into lcov ([1]). There hasn't yet been a release though. And the feature definitely has some rough edges, still. I'm planning to generate the 15->16 differential code coverage, once the feature freeze has been reached. Another nice thing provided by the in-development lcov is hierarchical output. I find that to be much easier to navigate than the current flat output, as e.g. used on coverage.postgresql.org. I've attached a screenshot showing the coverage differences of the backend between 15 and HEAD, as of a few days ago. "UNC" is uncovered new code, "LBC" is lost baseline coverage. I think for now it'd likely be a small script that'd generate the code coverage across versions. Do we want to have that in the source tree? Is there any interest in a) using the hierarchical output b) differential output on coverage.pg.o? Greetings, Andres Freund [1] https://github.com/linux-test-project/lcov/pull/169
Вложения
> On 4 Apr 2023, at 18:03, Andres Freund <andres@anarazel.de> wrote: > I'm planning to generate the 15->16 differential code coverage, once the > feature freeze has been reached. Cool! > I think for now it'd likely be a small script that'd generate the code > coverage across versions. Do we want to have that in the source tree? If it's published on pg.o like discussed below then I think it makes sense to include it. > Is there any interest in a) using the hierarchical output b) differential > output on coverage.pg.o? I would like to see that. If there are concerns about replacing the current coverage report with one from an unreleased (and potentially buggy) lcov, we could perhaps use diff.coverage.pg.o until it's released? -- Daniel Gustafsson
Hi, On 2023-04-04 09:03:45 -0700, Andres Freund wrote: > For quite a while I'd been wishing to see *differential* code coverage, to see > what changed in code coverage between two major releases. Unfortunately lcov > didn't provide that. A few months ago a PR for it has been merged into lcov > ([1]). There hasn't yet been a release though. And the feature definitely has > some rough edges, still. > > I'm planning to generate the 15->16 differential code coverage, once the > feature freeze has been reached. > > Another nice thing provided by the in-development lcov is hierarchical > output. I find that to be much easier to navigate than the current flat > output, as e.g. used on coverage.postgresql.org. I've generated the output for 15 vs HEAD, now that we're past feature freeze. Simpler version: https://anarazel.de/postgres/cov/15-vs-HEAD-01/ Version that tries to date the source lines using information from git: https://anarazel.de/postgres/cov/15-vs-HEAD-02/ The various abbreviations are explained (somewhat) if you hover over them. There's a few interesting bit of new code not covered by teests, that should easily be coverable: https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/bin/pg_dump/pg_dumpall.c.gcov.html#1014 https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/backend/utils/adt/array_userfuncs.c.gcov.html#675 https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/backend/utils/adt/array_userfuncs.c.gcov.html#921 https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/backend/utils/adt/pg_locale.c.gcov.html#1964 https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/backend/utils/adt/pg_locale.c.gcov.html#2219 https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/backend/parser/parse_expr.c.gcov.html#3108 https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/backend/nodes/nodeFuncs.c.gcov.html#1480 https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/backend/nodes/nodeFuncs.c.gcov.html#1652 https://anarazel.de/postgres/cov/15-vs-HEAD-01/src/backend/optimizer/util/plancat.c.gcov.html#348 https://anarazel.de/postgres/cov/15-vs-HEAD-01/contrib/postgres_fdw/connection.c.gcov.html#1316 https://anarazel.de/postgres/cov/15-vs-HEAD-01/contrib/postgres_fdw/deparse.c.gcov.html#399 Greetings, Andres Freund