Re: [Qemu-devel] [PATCH RFC v3 1/5] tests: New make target check-source

[Sascha Silbe <silbe@linux.vnet.ibm.com>]

Dear Markus,


I looked into the s390x specific failures as it would be nice to get
them fixed up right away. Fortunately it looks like all of them are
related to known shortcomings of the header check rather than actual
issues with the header files, so there's nothing for me to do. ;)


Markus Armbruster <armbru@redhat.com> writes:

[...]
> Some headers fail the test because they need to be compiled with
> -DNEED_CPU_H, and testing them is not yet implemented.  They are
> marked with a comment
>
>     /* NOTE: May only be included into target-dependent code */

The following header files either include target-s390x/cpu.h themselves
or expect it to have been included already so they should be marked with
the above note rather than the FIXME:

- hw/s390x/ipl.h
- hw/s390x/s390-pci-inst.h
- hw/s390x/s390-virtio.h


[linux-headers/asm-s390/kvm.h]
> +/* FIXME Does not pass make check-headers with CONFIG_WIN32, yet! */

Is this because of the linux/types.h include or is there some other
issue? Should we even check these imported headers?


[linux-user/host/*/hostdep.h]
> +/* FIXME Does not pass make check-headers, yet! */

The "struct ucontext" definition used by hostdep.h differs between host
architectures. Not sure if testing the headers for "foreign" host
architectures is possible; we should probably just skip them instead of
marking them with a FIXME.


Many or even most other non-softmmu target failures smell like
NEED_CPU_H.


[*/helper.h]
> +/* FIXME Does not pass make check-headers, yet! */

Probably never will. Looks like it's meant to be included only by some
magic preprocessor code.

Sascha
-- 
Softwareentwicklung Sascha Silbe, Niederhofenstraße 5/1, 71229 Leonberg
https://se-silbe.de/
USt-IdNr. DE281696641