[Help] Ask a question about lcov generating report

[W-M-R]

I used gdb to confirm that the spidrvr_unlink in the figure would run to this function, and then I performed gcov_dump, but the report showed that it did not run, which was quite accidental, and other files were normal. So I would like to ask you to see if you have encountered similar problems.

[henry2cox]

Yeah...we have seen all kinds of strange things :-)

A couple of questions:

• what toolchain and version? Upgrading to latest may resolve the issue.

• What happens if you use a different toolchain?
  If this is gcc: try LLVM. If LLVM, try gcc. Similarly: if using the llvm 'gcov path', then try the 'profdata path'. All of these alternatives are likely to produce slightly different results (and it is interesting to check those differences - e.g. with a differential coverage report). The differential report is often a source of useful clues.

• are you sure your code is hit? This looks like an async event handler. The coverage-instrumented code is going to be larger and slower than the optimized version - so races may come out differently. If you can write a log message from your spidrv_unlink callback and/or from the caller..see if you see a message from your testcase, using your optimized version and then your coverage build.

• is it possible that the GCDA file (runtime coverage data) for this particular testcase got written somewhere else, or didn't get written at all?

[W-M-R]

ok. thanks