Re: [PATCH v3 19/19] Add RISC-V to build-many-glibcs.py

[Joseph Myers <joseph at codesourcery dot com>]

On Tue, 26 Dec 2017, Palmer Dabbelt wrote:

> +        self.add_config(arch='riscv64',
> +                        os_name='linux-gnu',
> +                        glibcs=[{'variant': 'rv64imafdc-lp64d',
> +                                 'ccopts': '-march=rv64imafdc -mabi=lp64d',
> +                                 'cfg': ['--libdir=/usr/lib64/lp64d',
> +                                         'libc_cv_slibdir=/lib64/lp64d',
> +                                         'libc_cv_rtlddir=/lib']},

You shouldn't need any of those --libdir libc_cv_slibdir libc_cv_rtlddir 
settings in build-many-glibcs.py - they indicate something is missing in 
the port itself.  Specifically, you should have a LIBC_SLIBDIR_RTLDDIR 
call in an appropriate sysdeps configure.ac file or files.

Also, the use of lib<something>/<something> subdirectories will break the 
handling of --strip, which expects to be able to strip lib*/*.so and catch 
all the shared libraries that way.  Probably that code should change to 
something like '%s $(find %s/lib* -name "*.so")' in place of the present 
'%s %s/lib*/*.so', supposing that does work.

(Most RISC-V build-many-glibcs.py testing of course only needs to test 
building compilers and glibc for the configurations you're adding.  But 
for the --strip fix you should test at least one other configuration as 
well to make sure --strip continues to work in existing configurations 
with e.g. lib and lib64 directories.  There should be no need for any of 
your build-many-glibcs.py testing to use the full set of configurations 
for all architectures.)

-- 
Joseph S. Myers
joseph@codesourcery.com