https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115912
Bug ID: 115912 Summary: [15 regression] Harfbuzz testsuite fails (mvar_partial_instance test) since r15-1901-g98914f9eba5f19 Product: gcc Version: 15.0 Status: UNCONFIRMED Keywords: wrong-code Severity: normal Priority: P3 Component: rtl-optimization Assignee: unassigned at gcc dot gnu.org Reporter: sjames at gcc dot gnu.org CC: law at gcc dot gnu.org Target Milestone: --- Reproduction steps: ``` git clone https://github.com/harfbuzz/harfbuzz && cd harfbuzz CFLAGS="-O2 -march=znver2 -fno-vect-cost-model" CXXFLAGS="-O2 -march=znver2 -fno-vect-cost-model" meson setup build --wipe meson test -C build ``` Then can re-run just the bad test with adjusted paths as appropriate: ``` cd build LD_LIBRARY_PATH=/home/sam/git/harfbuzz/build/src /home/sam/git/harfbuzz/test/subset/run-tests.py util/hb-subset /home/sam/git/harfbuzz/test/subset/data/tests/mvar_partial_instance.tests ``` The difference is: ``` <!-- Most of this table will be recalculated by the compiler --> - <checkSumAdjustment value="0x5fad2240"/> + <checkSumAdjustment value="0xf5effee"/> - <mapping from="-0.49994" to="-0.62744"/> + <mapping from="-0.49994" to="-0.00006"/> - <mapping from="-0.24994" to="-0.2682"/> + <mapping from="-0.24994" to="-0.00006"/> ``` It looks suspicious that both values go to -0.00006. I'll narrow it down more now.