Hi Paul,
That is very well done. Thanks for picking it up and running with it. OK on both the fix and the dumping of the gsymbols.
Committed, thanks.
You might consider back porting both this patch and my fix for the original bug to 9-branch.
Both patches apply cleanly together on gcc-9 (well, except for the
whitespace fix of a comment that didn't exist yet in gcc-9, so
I think we can ignore than), and pass regression-tests on POWER9.
So, I will commit this tonight.
Regards
Thomas
