[Tests] Include branch coverage info in coverage test #10511
Conversation
|
Here is a possible hack to force a variable list of arguments to be treated as 'used' for the purpose of static inline void MarkUsed() {}
template<typename T, typename... Args> static inline void MarkUsed(const T& t, const Args&... args)
{
(void)t;
MarkUsed(args...);
}
#ifdef USE_COVERAGE
#define LogPrintf(...) do { MarkUsed(__VA_ARGS__); } while(0)
#define LogPrint(category, ...) do { MarkUsed(__VA_ARGS__); } while(0)
#else |
|
Used @sipa's hack to shut up the warnings. |
|
Cool. Concept ACK. |
Added an option to configure to allow for branch coverage statistics gathering. Disabled logprint macro when coverage testing is on so that unnecessary branches are not analyzed.
contrib/filter-lcov.py
Outdated
| with open(tracefile, 'r') as f: | ||
| with open(outfile, 'w') as wf: | ||
| for line in f: | ||
| if pattern in line: |
There was a problem hiding this comment.
Perhaps instead use line.startswith("SF:" + pattern) instead? That would prevent matching on variables that accidentally match the pattern name.
contrib/filter-lcov.py
Outdated
| in_remove = True | ||
| if not in_remove: | ||
| wf.write(line) | ||
| if 'end_of_record' in line: |
There was a problem hiding this comment.
Perhaps use line == 'end_of_record' instead? Same reason as above.
achow101
force-pushed
the
lcov
branch
2 times, most recently
from
8687849
to
405b86a
Compare
|
utACK 405b86a |
| @@ -0,0 +1,24 @@ | |||
| #!/usr/bin/env python3 | |||
There was a problem hiding this comment.
explicit license is missing:
# Copyright (c) 2017 Bitcoin Core Developers
# Distributed under the MIT software license, see the accompanying
# file COPYING or http://www.opensource.org/licenses/mit-license.php.| with open(outfile, 'w') as wf: | ||
| for line in f: | ||
| if line.startswith("SF:") and pattern in line: | ||
| in_remove = True |
There was a problem hiding this comment.
Not sure why you need the var in_remove. Couldn't you write wf.write(line);continue; here?
There was a problem hiding this comment.
The format of the file is
SF:<path to file>
<bunch of lines>
end_of_record
<repeat>
In order to know that we are in a <bunch of lines> section that should be removed (because the SF:... line matches our removal pattern, we need to have a variable to remember that we shouldn't be writing these lines to the output file. That is what in_remove is for.
Instead of using lcov -r (which is extremely slow), first use a python script to perform bulk cleanup of the /usr/include/* coverage. Then use lcov -a to remove the duplicate entries. This has the same effect of lcov -r but runs significantly faster
|
The lcov stuff was added ages ago, when lcov's --no-external was too new to rely on. I think it'd be ok now to require >=1.10 and forego all of the filtering. |
|
@achow101 Are you still working on this? |
|
@MarcoFalke Yes |
|
Would it be okay with everyone if I just merged #10565 into this PR? It all falls under "improvements to coverage data". |
|
Needs rebase. |
|
Closing this as #10565 contains this PR (I totally forgot) except for the copyright header I added last commit :/ |
This PR adds an option to configure to include branch coverage statistics be gathered during
make cov. It also disables logging when coverage is enabled so that the logging branches are not included in the coverage stats. This has a side effect of causing-Wunused-variablewarnings during compile time when coverage is enabled.