https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64583
--- Comment #7 from Markus Trippelsdorf <trippels at gcc dot gnu.org> --- (In reply to Jan Hubicka from comment #6) > In the testcase the symbol in question seems to be optimized out because it > is comdat and unused. That looks valid. Does possibly fix for PR64612 help > here? Martin has got a more detailed linker error: http://pastebin.com/wbcKdpVA So _ZN5blink10FontFamilyD1Ev seems to be the culprit. Will test the PR64612 patch tomorrow.