ILP32 for ARM64: testing with glibc testsuite

[Steve Ellcey]

On Wed, 2016-11-16 at 15:22 +0400, Maxim Kuvyrkov wrote:
> > 
> > On Nov 9, 2016, at 1:56 PM, Yury Norov <ynorov at caviumnetworks.com>
> > wrote:
> > 
> > > 
> > > Below is the results of glibc testsuite run for aarch64/lp64

I have been running the glibc testsuite as well.  I have only run it on
an ILP32 enabled kernel.  Using that kernel, top-of-tree glibc, and the
ILP32 glibc patches I have no LP64 regressions.  There are 5 failures
in LP64 mode but I get them with vanilla top-of-tree glibc sources too.
They are:
	nptl/eintr1 (I actually don't run this because it kills the 'make check')
	debug/tst-backtrace5
	debug/tst-backtrace6
	nptl/tst-stack4
	nptl/tst-thread_local1

In ILP32 mode I get 33 failures, they include the above failures (minus
nptl/tst-thread_local1) plus:

	c++-types-check
	conform/ISO11/inttypes.h/conform
	conform/ISO11/stdint.h/conform
	conform/ISO99/inttypes.h/conform
	conform/ISO99/stdint.h/conform
	conform/POSIX2008/inttypes.h/conform
	conform/POSIX2008/stdint.h/conform
	conform/XOPEN2K/inttypes.h/conform
	conform/XOPEN2K/stdint.h/conform
	conform/XOPEN2K8/inttypes.h/conform
	conform/XOPEN2K8/stdint.h/conform
	elf/tst-tls1
	elf/tst-tls1-static
	elf/tst-tls2
	elf/tst-tls2-static
	elf/tst-tls3
	math/check-abi-libm
	math/test-double
	math/test-double-finite
	math/test-float
	math/test-float-finite
	misc/tst-sync_file_range
	nptl/tst-cancel26
	nptl/tst-cancel27
	nptl/tst-sem3
	rt/tst-mqueue1
	rt/tst-mqueue2
	rt/tst-mqueue4
	rt/tst-mqueue7
	stdlib/tst-makecontext3

I am currently looking at these ILP32 regressions (starting with the
tls failures) to see if I can figure out what is happening with them.

Steve Ellcey
sellcey at caviumnetworks.com