No Linux kernel code coverage when using GCC 15

[matttbe]

Hello,

I recently switched to GCC 15, and since then, I have no more code coverage.

# lcov --capture --keep-going -j 2 --ignore-errors mismatch --rc geninfo_unexecuted_blocks=1 --include /net/mptcp/ --function-coverage --branch-coverage -b /workspace/mptcp_net-next/.virtme/build -o /workspace/mptcp_net-next/.virtme/results/0d7b92893723/normal/kernel.lcov
Auto-detecting gcov kernel support.
Found upstream gcov kernel support at /sys/kernel/debug/gcov
Copying data to temporary directory /tmp/lyaNu17iw7
Capturing coverage data from /tmp/lyaNu17iw7
geninfo cmd: '/usr/bin/geninfo /tmp/lyaNu17iw7 --toolname lcov --output-filename /workspace/mptcp_net-next/.virtme/results/0d7b92893723/normal/kernel.lcov --keep-going --base-directory /workspace/mptcp_net-next/.virtme/build --ignore-errors mismatch --parallel 2 --branch-coverage --rc geninfo_unexecuted_blocks=1 --include /net/mptcp/'
Found gcov version: 15.2.0
Using intermediate gcov format
Recording 'internal' directories:
        /tmp/lyaNu17iw7
        /workspace/mptcp_net-next/.virtme/build
Writing temporary data to /tmp/geninfo_dat7lqj
Scanning /tmp/lyaNu17iw7 for .gcda files ...
Found 18 data files in /tmp/lyaNu17iw7
using: chunkSize: 7, nchunks:3, intervalLength:0
lcov: WARNING: (mismatch) "/workspace/mptcp_net-next/net/mptcp/protocol.h":594: mismatched exception tag for id 4, 4: '0' -> '1'
        (use "lcov --ignore-errors mismatch,mismatch ..." to suppress this warning)
lcov: WARNING: (mismatch) "/workspace/mptcp_net-next/net/mptcp/protocol.h":600: mismatched exception tag for id 0, 0: '0' -> '1'
lcov: WARNING: (mismatch) "/workspace/mptcp_net-next/net/mptcp/protocol.h":594: mismatched exception tag for id 4, 4: '0' -> '1'
lcov: WARNING: (mismatch) "/workspace/mptcp_net-next/net/mptcp/protocol.h":600: mismatched exception tag for id 0, 0: '0' -> '1'
Finished processing 18 GCDA files
Apply filtering..
Finished filter file processing
Finished .info-file creation
Summary coverage rate:
  source files: 20
  lines.......: 0.0% (0 of 8381 lines)
  functions...: 0.0% (0 of 548 functions)
  branches....: 0.0% (0 of 6085 branches)
Message summary:
  4 warning messages:
    mismatch: 4
Removing temporary directories.

See the 0.0%. Here is the created file: kernel.lcov.gz

There are more examples of .lcov files produced there if needed: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/18831360999 (in the artifacts, starting with lcov-)

I'm using lcov v2.3.1, but I have the same issue with v2.3.2.

I'm not exactly sure what I should look at. The .gcda and .gcno files are not empty:

# cat /sys/kernel/debug/gcov/workspace/mptcp_net-next/.virtme/build/net/mptcp/token_test.gcda | wc -c
16
# cat /sys/kernel/debug/gcov/workspace/mptcp_net-next/.virtme/build/net/mptcp/token_test.gcno | wc -c
15696

If you need more captured files, here is a build with GCC 14, and one with GCC 15. You can see the 0.0% in the summary, and the capture files are in the artifacts, starting with lcov-.

[oberpar]

Thank you for the report. The reason for this error is a missing compatibility patch in the Linux kernel. These patches are necessary whenever GCC introduces new GCOV-Counter types, such as happened with GCC 15 with this GCC commit.

To fix this, the value of GCOV_COUNTERS needs to be adjusted to 10 for GCC versions starting with 15, similar to how this was fixed the last time via this kernel commit, so something like the following patch should do the trick:

diff --git a/kernel/gcov/gcc_4_7.c b/kernel/gcov/gcc_4_7.c
index a08cc076f332..ffde93d051a4 100644
--- a/kernel/gcov/gcc_4_7.c
+++ b/kernel/gcov/gcc_4_7.c
@@ -18,7 +18,9 @@
 #include <linux/mm.h>
 #include "gcov.h"
 
-#if (__GNUC__ >= 14)
+#if (__GNUC__ >= 15)
+#define GCOV_COUNTERS                  10
+#elif (__GNUC__ >= 14)
 #define GCOV_COUNTERS                  9
 #elif (__GNUC__ >= 10)
 #define GCOV_COUNTERS                  8

Could you give this a try?

[matttbe]

@oberpar thank you for your reply, and the explanations! (Next time, I know where to look!)

Indeed, that fixes the issue on my side, I no longer have 0.0%. I will re-run the full test suite with this patch, and report back if I notice anything wrong.

I guess you are going to send a patch to fix that, right?

[matttbe]

Oh, sorry, I didn't know my commit in a completely different repo would close this issue!

Feel free to re-open it if preferred: it looks like I don't have the permissions to do that.

[matttbe]

I confirm the patch seems to fix the issue: https://ci-results.mptcp.dev/lcov/export/index.html

[oberpar]

Thank you for testing. Yes, I'll send the patch to LKML. To attribute Reported-by and Tested-by to you, is Matthieu Baerts <matttbe@kernel.org> the correct address for that?

[matttbe]

Thank you!

Yes, the address is correct.

[matttbe]

Thank you for having sent the patch!

[oberpar]

You're welcome! And thanks again for reporting and testing.

For reference, here's the associated LKML post:

https://lore.kernel.org/all/20251028115125.1319410-1-oberpar@linux.ibm.com/