We have coverage targets in our Makefile for using gcov to display line
coverage based on our test suite. The way I like to do it is to run:
make coverage-test
make coverage-report
This leaves the repo in a state where every X.c file that was covered has an
X.c.gcov file containing the coverage counts for every line, and "#" at
every uncovered line.
There have been a few bugs in recent patches what would have been caught if
the test suite covered those blocks (including a few of mine). I want to
work towards a "sensible" amount of coverage on new topics. In my opinion,
this means that any logic should be covered, but the 'die()' blocks in error
cases do not need to be covered.
It is important to not measure the coverage of the codebase by what old code
is not covered. To help, I created the 'contrib/coverage-diff.sh' script.
After creating the coverage statistics at a version (say, 'topic') you can
then run
contrib/coverage-diff.sh base topic
to see the lines added between 'base' and 'topic' that are not covered by
the test suite. For example, I ran this against the 'next' branch (22e244b)
versus 'master' (150f307) and got the following output:
fsck.c
fb8952077df (René Scharfe 2018-09-03 14:49:26 + 212)
die_errno("Could not read '%s'", path);
list-objects-filter-options.c
f12b8fc6d3b (Matthew DeVore 2018-09-13 17:55:27 -0700 56)
if (errbuf) {
f12b8fc6d3b (Matthew DeVore 2018-09-13 17:55:27 -0700 57)
strbuf_init(errbuf, 0);
f12b8fc6d3b (Matthew DeVore 2018-09-13 17:55:27 -0700 58)
strbuf_addstr(
f12b8fc6d3b (Matthew DeVore 2018-09-13 17:55:27 -0700 62)
return 1;
list-objects-filter.c
77d7a65d502 (Matthew DeVore 2018-09-13 17:55:26 -0700 47)
BUG("unknown filter_situation: %d", filter_situation);
f12b8fc6d3b (Matthew DeVore 2018-09-13 17:55:27 -0700 100)default:
f12b8fc6d3b (Matthew DeVore 2018-09-13 17:55:27 -0700 101)
BUG("unknown filter_situation: %d", filter_situation);
77d7a65d502 (Matthew DeVore 2018-09-13 17:55:26 -0700 152)
BUG("unknown filter_situation: %d", filter_situation);
77d7a65d502 (Matthew DeVore 2018-09-13 17:55:26 -0700 257)
BUG("unknown filter_situation: %d", filter_situation);
77d7a65d502 (Matthew DeVore 2018-09-13 17:55:26 -0700 438)
BUG("invalid list-objects filter choice: %d",
list-objects.c
f447a499dbb (Matthew DeVore 2018-08-13 11:14:28 -0700 197)
ctx->show_object(obj, base->buf, ctx->show_data);
ll-merge.c
d64324cb60e (Torsten Bögershausen 2018-09-12 21:32:02 +0200 379)
marker_size = DEFAULT_CONFLICT_MARKER_SIZE;
midx.c
56ee7ff1565 (Derrick Stolee 2018-09-13 11:02:13 -0700 949)
return 0;
cc6af73c029 (Derrick Stolee 2018-09-13 11:02:25 -0700 990)
midx_report(_("failed to load pack-index for packfile %s"),
cc6af73c029 (Derrick Stolee 2018-09-13 11:02:25 -0700 991)
e.p->pack_name);
cc6af73c029 (Derrick Stolee 2018-09-13 11:02:25 -0700 992)
break;
remote-curl.c
c3b9bc94b9b (Elijah Newren 2018-09-05 10:03:07 -0700 181)
options.filter = xstrdup(value);
submodule.c
df255b8cac7 (Brandon Williams 2018-08-08 15:33:22 -0700 1738)
die(_("could not create directory '%s'"), new_gitdir.buf);
Using this 'git blame' output, we can quickly inspect whether the uncovered
lines are appropriate. For instance:
1. The line in builtin/commit.c is due to writing the commit-graph file
when GIT_TEST_COMMIT_GRAPH is enabled, which is not on by default in the
test suite. Being uncovered is expected here.
2. The lines in builtin/worktree.c are all related to error conditions.
This is acceptable.
3. The line in builtin/rev-list.c is a flag replacement in a block that is
otherwise unchanged. It must not be covered by the test suite normally.
This could be worth adding a test to ensure the new logic maintains old
behavior.
4. The lines in read-cache.c are part of a new block for the condition "if
(expand_name_field)" as part of an optimization. These lines should
probably be covered before that series is merged to 'next'. I understand
that Ben and Duy are continuing work in this direction [1].
I used this approach for 'next' over 'master' and got a larger list, some of
which I have already submitted tests to increase coverage [2] or will be
covered by topics not in 'next' [3].
Thanks, -Stolee
CHANGES IN V3: I took Junio's perl script verbatim, which speeds up the
performance greatly. Some of the other sed commands needed some massaging,
but also added extra cleanup. Thanks for the help!
[1]
https://pu