[oe] [meta-oe][PATCH] stress-ng: Upgrade from 0.09.37 to 0.09.45

[Randy MacLeod]

On 11/12/18 4:14 PM, Randy MacLeod wrote:
> Drop the upstreamed patch:
>     0001-stress-ng-don-t-enable-TARGET_CLONES-for-musl.patch
> Rebase other patches except for the stress-malloc.c change
> which has made it upstream:
>     c3a2e662 stress-malloc: use HAVE_MALLOPT check for mallopt glibc function
> 
> Signed-off-by: Randy MacLeod <Randy.MacLeod at windriver.com>
> ---
>   ...force-sync-after-build-in-case-reboo.patch | 21 +++++----
>   ...01-Several-changes-to-fix-musl-build.patch | 47 ++++++++-----------
>   ...-don-t-enable-TARGET_CLONES-for-musl.patch | 30 ------------
>   ...ess-ng_0.09.37.bb => stress-ng_0.09.45.bb} |  6 +--
>   4 files changed, 34 insertions(+), 70 deletions(-)
>   delete mode 100644 meta-oe/recipes-test/stress-ng/stress-ng/0001-stress-ng-don-t-enable-TARGET_CLONES-for-musl.patch
>   rename meta-oe/recipes-test/stress-ng/{stress-ng_0.09.37.bb => stress-ng_0.09.45.bb} (78%)

...

> +SRC_URI[md5sum] = "b03744c2eb68bf7e9a300e78e397f348"
> +SRC_URI[sha256sum] = "0741e3004bf590bb7af3db979a46fe89bee7aaad6065cd1d87d0b7fa49046cb2"
>   
>   UPSTREAM_CHECK_URI ?= "http://kernel.ubuntu.com/~cking/tarballs/${BPN}/"
>   UPSTREAM_CHECK_REGEX ?= "(?P<pver>\d+(\.\d+)+)\.tar"
> 


Builds fine for all qemu machines for glibc and musl,
except for qemuarm/musl which fails to build due to:

| CC stress-nop.c
| CC stress-null.c
| stress-fp-error.c: In function 'stress_fp_error':
| stress-fp-error.c:98:10: error: 'FE_INVALID' undeclared (first use in 
this function); did you mean 'EINVAL'?
|     EDOM, FE_INVALID);
|           ^~~~~~~~~~
|           EINVAL
| stress-fp-error.c:98:10: note: each undeclared identifier is reported 
only once for each function it appears in
| stress-fp-error.c:104:12: error: 'FE_DIVBYZERO' undeclared (first use 
in this function); did you mean 'FP_ZERO'?
|     ERANGE, FE_DIVBYZERO);
|             ^~~~~~~~~~~~
|             FP_ZERO
| CC stress-oom-pipe.c
| stress-fp-error.c:139:7: error: 'FE_INEXACT' undeclared (first use in 
this function); did you mean 'AF_INET'?
|     0, FE_INEXACT);
|        ^~~~~~~~~~


I'm not sure how to handle that since in the source file,
there does seem to be a guard already:

#if defined(EDOM)
                 stress_fp_clear_error();
                 stress_fp_check(args, "log(-1.0)", log(-1.0), NAN,
                         EDOM, FE_INVALID);
#endif

but clearly EDOM is defined for qemuarm.
Should I stub out the whole body of the function:
stress_fp_error() if FE_INVALID isn't defined?

Any better ideas?


Oh and stress-ng runs fine for both glibc/musl in qemux86-64.

../Randy



-- 
# Randy MacLeod
# Wind River Linux