[Bug driver/85142] Wrong -print-multi-os-directory & -print-multi-lib output for riscv64 + multilib

["joseph at codesourcery dot com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85142

--- Comment #1 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
On Sat, 31 Mar 2018, david.abdurachmanov at gmail dot com wrote:

> GCC 7.3.1 is available in Fedora RISC-V stage4 and is configured with multilib,
> but only one ABI is selected. This is done to provide multilib paths, which are
> required by glibc.
> 
> # gcc -print-multi-directory               
> .                        
> # gcc -print-multi-lib
> .;
> lib64/lp64d;@march=rv64imafdc@mabi=lp64d

If there is only one multilib, there should only be one line in the 
-print-multi-lib output.  So something seems wrong here (whether a mistake 
in how the compiler was configured - you haven't provided the detailed 
configure options and confirmed reproducing the issue with those options 
with unmodified upstream GCC - or a bug in the RISC-V t-* files).

> # gcc -print-multi-os-directory
> .
> 
> Based on GCC man page -print-multi-os-directory should return ../lib64/lp64d

Yes, if that's where libraries such as libstdc++ are installed by GCC.

> Instead of:
> 
>     lib64/lp64d;@march=rv64imafdc@mabi=lp64d
> 
> We should have:
> 
>     ../lib64/lp64d;@march=rv64imafdc@mabi=lp64d

No, that's wrong.  For a multilib-disabled compiler you should have ".;" 
and nothing else in the -print-multi-lib output.

Directory names in -print-multi-lib / -print-multi-directory output are 
the *non-OS* directory names, which are used to install libgcc (and crt*.o 
etc.) copies in lib/gcc/$target/$version.  They must never contain a ".." 
component because that would mean the $version directory component getting 
lost.

> All the paths seem to be related to "some lib directory" (/usr/lib) according
> to man page.

Only the -print-multi-os-directory names use *OS* conventions, meaning 
relative to a directory called lib.