--cppcheck-build-dir not helping

[Zach Thompson]

I'm new to cppcheck so I may just be using it wrong, but using a build directory isn't saving any time for me - using both the GUI and command line. Even if I run the same command back to back without changing any files, it takes the same time (approximately 2 minutes for 190 files)

My command line looks like --cppcheck --cppcheck-build-dir=build-dir --project=project.sln "--project-configuration=RelWithDebInfo|x64" --enable=all -j8

Any thoughts?

[CHR]

Do the files in build-dir also change with every execution?

[Zach Thompson]

Yes, they get completely refreshed.

[CHR]

The decision to run the analysis again happens here: https://github.com/danmar/cppcheck/blob/e17af6d2d865def4f3557bc374bd9892cd943709/lib/cppcheck.cpp#L749
Maybe you can look at the hashed strings?

[Zach Thompson]

What am I looking for exactly? About 80% of the files created in build-dir are .a1 and have their its own unique analyzerinfo hash number but are otherwise identical and look like:

<?xml version="1.0"?>
<analyzerinfo hash="8543757348695670605">
        <error id="preprocessorErrorDirective" severity="error" msg="failed to expand &apos;BOOST_MPL_AUX_NA_SPEC&apos;, Invalid ## usage when expanding &apos;BOOST_PP_TUPLE_ELEM_II&apos;: Unexpected token &apos;)&apos;" verbose="failed to expand &apos;BOOST_MPL_AUX_NA_SPEC&apos;, Invalid ## usage when expanding &apos;BOOST_PP_TUPLE_ELEM_II&apos;: Unexpected token &apos;)&apos;">
            <location file="C:\samepatheverytime\file.cpp" line="29" column="0"/>
        </error>
</analyzerinfo>

There are also about 10% of .snanalyzerinfo files as well but it seems like they are only generated when their associated .a1 file included a FileInfo check error such as Uninitialized variables or CheckUnusedFunction

[CHR]

You could dump the toolinfo string to a file (or look at it in the debugger) and compare the content from different runs to see if they are different. I remember an issue where suppressions changed randomly due to missing synchronization between threads.

[Zach Thompson]

Hash numbers remain the same after each run - however they change depending on whether I run it via command line or the GUI. Another strange thing, if I use --enable=all then the files are not overwritten but I still don't get any improved speed

[CHR]

What happens in this line? https://github.com/danmar/cppcheck/blob/e17af6d2d865def4f3557bc374bd9892cd943709/lib/cppcheck.cpp#L764
Is the analysis skipped when nothing has changed?
Are you on Windows?