[RFC] Remove special flags of libc.5.so and libc.4.so

[Adhemerval Zanella]

On 02/03/2022 16:37, Raoni Fassina Firmino wrote:
> On Tue, Jan 11, 2022 at 04:22:43PM -0300, AL glibc-alpha wrote:
>>
>>
>> On 22/12/2021 12:31, Lucas A. M. Magalhaes via Libc-alpha wrote:
>>> I've tested it in a AT build and was able to build and link programs as
>>> expected. For other archs I try with build-many-glibcs.py, all
>>> succeeded.
>>>
>>> Any other suggestions on how to test this?
>>>
>>> Florian am I missing something? Should we add a restriction to cache
>>> just new libc6.so compatible libraries?
>>>
>>> -- 8< --
>>>
>>> The older libcs version are obsolete for over twenty years now. This
>>> commit removes special flags of libc.5.so and libc.4.so.. It assume that
>>> all libraries cached are libc.6.so compatible and wrote FLAG_ELF_LIBC6.
>>>
>>> Tested on x86_64, powerpc, powerpc64 and powerpc64le.
>>> Tested with build-many-glibcs.py for the other affected architectures.
>>
>> I think it makes sense to stop support libc.4 and libc.5 and assume
>> FLAG_ELF_LIBC6 by ldconfig.
>>
>> I also think it should be safe to just get rid of the arch-specific
>> SYSDEP_KNOWN_INTERPRETER_NAMES, at least in theory it should not be
>> used by the cache (so there is not need to actually classify it).
> 
> Lucas's RFC already removed it for a bunch of architectures, only
> leaving i386, aarch64 and mips64, and If my grep did not failed me It
> seems like the only uses of YSDEP_KNOWN_LIBRARY_NAMES and
> SYSDEP_KNOWN_INTERPRETER_NAMES were in known_libs and interpreters
> respectively, both removed in the RFC, so it seems that definitions can
> be safely removed (in the context of the RFC).
> 
> Not sure if there is a need for more opinions or consensus about this
> RFC's idea and approach to move forward with a proper patch for review.
> 
> 

I think the rationale of the patch is ok, it just need the the
SYSDEP_KNOWN_LIBRARY_NAMES and SYSDEP_KNOWN_INTERPRETER_NAMES
cleanup for aarch64, i386, and mips64.