On Tue, 19 Feb 2019, Martin Liška wrote: > Sure, fixed in attached patch. This version is OK with the indentation in math/Makefile fixed to match the surrounding lines, presuming x86_64 and x32 are indeed mutually exclusive conditions for the purpose of this multilib selection. *But* there could be potential interactions with Zack's patch to tokenize installed headers to check for obsolete typedefs, since that of course is tokenizing headers as C and this is adding a Fortran header to the installed headers. (The existing check-installed-headers tests skip bits/* so wouldn't be affected.) So maybe we should have a way to distinguish Fortran headers, or maybe no problems arise in practice from tokenizing this one as C and the new Python test could change to skip such headers (it already has checks for particular header names that shouldn't be checked) if they cause problems in future. -- Joseph S. Myers joseph@codesourcery.com