Thanks for the report.

This appears to be a false positive and unrelated to patch 141473.

All six checks of the new pr126348 regression test pass in the reported AArch64 build. The reported gcc.misc-tests/gcov-39.c test is compiled with --coverage -fpath-coverage and without -flto, so it does not exercise lto_write_tree_1, which is the only compiler implementation code changed by this patch.

Moreover, the identical 1/7-versus-1/8 failure occurs in nine consecutive AArch64 precommit builds (10755-10763), covering several unrelated patches and patch series, including C++ contracts changes and RISC-V-only backend changes unrelated to AArch64 code generation. The reference result reports 51 total prime paths, while the precommit result reports 50, suggesting a reference/precommit baseline or environment discrepancy rather than an effect of this patch.

Could you please disregard the regression attribution to this patch and investigate or update the baseline/flaky-test filtering as appropriate?

Current build: http://54.172.246.49:9090/jobs/tcwg_gcc_check--master-aarch64-precommit/buil...

Reference build: http://54.172.246.49:9090/jobs/tcwg_gcc_check--master-aarch64-build/builds/1...

First build in the consecutive sequence: http://54.172.246.49:9090/jobs/tcwg_gcc_check--master-aarch64-precommit/buil...

RISC-V-only example at the other end of the sequence: http://54.172.246.49:9090/jobs/tcwg_gcc_check--master-aarch64-precommit/buil...