[OE-core] ptest with BBCLASSEXTEND

[richard.purdie at linuxfoundation.org]

On Mon, 2019-06-03 at 11:10 +0800, Changqing Li wrote:
> Based on this list, except lib*perl and packages that are dealing
> with by others,
> 
> I picked some for your reference to evaluate the workload.  If more
> info is needed,
> 
> feel free to tell me.

If it helps, I reran the minimal image comparison and found the
following issues:

Recipes where we have problems (differences) in minimal images:

bash                         | 76          | 5       | 0         | 163
diffutils                    | 19          | 0       | 2         | 1
e2fsprogs                    | 0           | 0       | 0         | 111
elfutils                     | 194         | 0       | 9         | 32
gettext                      | 354         | 21      | 37        | 32
gzip                         | 19          | 0       | 3         | 3
kbd                          | 5           | 2       | 0         | 0
m4                           | 145         | 7       | 18        | 5
openssh                      | 15          | 1       | 1         | 145
parted                       | 41          | 1       | 22        | 5
perl                         | 2413        | 18      | 245       | 861
python                       | 301         | 6       | 41        | 327
python3                      | 30277       | 2       | 1043      | 1012

Others:

acl tests are timing out?

gawk shows failures in sdk image but minimal:
gawk                         | 407         | 24      | 0         | 20

gstreamer shows two fewer failures in minimal image:
gstreamer1.0                 | 104         | 1       | 0         | 104

(I'm using 
https://autobuilder.yocto.io/pub/non-release/20190601-2/testresults/testresult-report.txt
for comparison)

There are also some open ptest bugs in bugzilla which I didn't
duplicate here.

The above list is much smaller than the original particularly now we
have the perl module issues resolved.

Cheers,

Richard

> diffutils   depend on perl
> elfutils:   mingli have work on it and send patch to oe-core mail
> list
> gettext:  depend on certain locales for some tests
> 
>                depend on autoconf for some test
> gzip:       have tests depend on diff, while diff in busybox 
> incompatible with POSIX
>                have tests depend on grep, while grep in busybox
> don't 
> support context options
>                have tests depend on perl
> kbd:       have tests depend on readlink,  busybox's readlink don't 
> support -e option
> m4:       one test depend on perl, some test skip since depend on 
> certain locale/codeset, need to check one by one
> 
>