https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89260
--- Comment #7 from Martin Liška <marxin at gcc dot gnu.org> --- (In reply to ohaiziejohwahkeezuoz from comment #5) > Reproducing this might be somewhat complicated, because harfbuzz has some > dependencies, which would also needed to be built. But I have captured the > failing command, and can run gdb on it interactively after segault, so if > you have any suggestions on what to look for, or how to extract more detail > other than the backtrace I've provided, I can do some debugging on my side. > > It's failing at the final linking step (probably during the load/parsing of > lto object files from disk), where all the many object files are processed > by the LTO, so I can't provide the reduced testcase. > > I'll check if harfbuzz 2.3.1 builds with gcc 8.2.0 (2.3.0 did, so that might > help in reducing the search space). I would like to get a reproducible test-case. You can run the linking command line with --save-temps --verbose. And then you can reduce number of objects needed. For .a files, one would need to extract them and provide .o files on command line. Doing that one can get minimal set of objects. And then you can create pre-processed files for these (-E option). It will generate a minimal reproducible test-case.
