[patches] Re: RISC-V glibc port, v4

[Darius Rad]

On 01/15/2018 11:13 AM, Joseph Myers wrote:
> On Sat, 13 Jan 2018, Palmer Dabbelt wrote:
> 
>> It's been a bit more than two weeks since the last patch set, and I think we're
>> starting to get something that's in reasonable shape.  Thanks to Darius for
>> putting in a lot of time into the v4.  We've also gotten the test suite in a
>> bit better shape: we're down to 200 failures when running natively on Linux,
>> and all the test suite failures when running in user mode emulate appear to be
>> environment issues.  As far as I know the only remaining issue in our port is
>> to triage the test suite failures, most of which we're hoping are just
>> deficiencies in our current environment.  Of course, now that I've said that
>> I'm sure we'll find a bunch of bugs all over the place :)
> 
> Although no more than about 20 architecture-specific failures (for each 
> configuration tested) should be expected for a port in reasonable shape 
> for inclusion, even before that point it makes sense to provide the list 
> of failures, as some failures may point reviewers to specific issues with 
> the port - details of test failures are necessary information for a full 
> review of a port.  (Even better, also make the "make check" output and the 
> .out files for failed tests available somewhere.)
> 

Summary of test results:
    159 FAIL
   5413 PASS
      4 UNRESOLVED
     21 UNSUPPORTED
     18 XFAIL

I have placed output here [1], including .out and .out-result for the
failing (in fact, non-PASS) tests.  This was run with qemu-system on
RV64 LP64D.  This also reflects two small changes on top of the V4 patches.

About 35 tests fail because they attempt to load incompatible libraries
from /lib, rather than the libraries under test.  These error messages
are not reflected in the test output, but are in the output of 'make
check' (output.log in the tar file).

[1] http://bluespec.com/tmp/glibc-check.tar.xz

>> Thanks to everyone who has helped review and contributed to our port.  Like
>> usual I believe we've taken everything into account (and hopefully Darius' hit
>> rate is a bit better than mine), but if we've missed anything feel free to ping
>> it and we'll take another look.
> 
> * This version is still missing leading __ on the parameter names for 
> __riscv_flush_icache in sys/cachectl.h.
> 
> * Of the pieces of multi-ABI support I mentioned in 
> <https://sourceware.org/ml/libc-alpha/2018-01/msg00008.html>, you're 
> missing dl-cache.h and an ldd_rewrite_script setting.  (To test all this 
> you need to set up a system with libraries for multiple ABIs installed in 
> the appropriate directories, run ldconfig and make sure that the libraries 
> for the different ABIs are properly listed in the cache with the right 
> flags set on them - and that ldd installed with glibc for one ABI works 
> properly on binaries for both that and the other ABI.  The glibc testsuite 
> probably doesn't cover any of this.)
> 
> * You're missing sysdeps/riscv/nofpu/Implies pointing to ieee754/soft-fp 
> (or more nofpu directories than that, if a single such Implies file 
> doesn't suffice to get ieee754/soft-fp before the other ieee754 
> directories in the sysdeps directory ordering for soft-float 
> configurations).
> 
> * As I said in 
> <https://sourceware.org/ml/libc-alpha/2017-06/msg00645.html>, you should 
> not have a riscv/soft-fp directory.  Previous versions got this right, 
> this one appears to have gone backwards.
> 
>> * Copyright year has been updated to 2018, and "contributed by" lines have been
>>   removed.
> 
> But you have one in sysdeps/unix/sysv/linux/riscv/readelflib.c (which is 
> also missing an explanatory comment before the copyright notice).
>