On 04/17/2018 02:20 PM, Marek Marczykowski-Górecki wrote:
On Tue, Apr 17, 2018 at 02:00:25PM -0600, Jim Fehlig wrote:
> Your response in the V6 thread about "conflicting types (with
> libxl_get_free_memory_0x040700)" was a good hint. Things work fine for me on
> Xen 4.4 through 4.10 with the following squashed in
>
> diff --git a/tests/Makefile.am b/tests/Makefile.am
> index 2c0d1311c..8c4b6c220 100644
> --- a/tests/Makefile.am
> +++ b/tests/Makefile.am
> @@ -528,6 +528,7 @@ libxlxml2domconfigtest_LDADD = $(libxl_LDADDS) $(LIBXML_LIBS)
>
> virmocklibxl_la_SOURCES = \
> virmocklibxl.c
> +virmocklibxl_la_CFLAGS = $(LIBXL_CFLAGS)
> virmocklibxl_la_LDFLAGS = $(MOCKLIBS_LDFLAGS)
> virmocklibxl_la_LIBADD = $(MOCKLIBS_LIBS)
>
> But I'm still confused as to how this works for you. Any idea about that? :-)
Oh, I see it now. For me, -DLIBXL_API_VERSION=0x040400 landed in CFLAGS,
not only LIBXL_CFLAGS. According to config.log it happened just after
testing for xenlight - first, the test using pkg-config failed (something
wrong with my libxl installation), but then fallback to manual check.
And since then, every gcc call have -DLIBXL_API_VERSION=0x040400.
Looking at m4/virt-driver-libxl.m4, I think it's because saved
old_CFLAGS is overridden by LIBVIRT_CHECK_LIB with actual CFLAGS (now
containing -DLIBXL_API_VERSION=0x040400).
Ah, looks to be the case. But that is existing and can be fixed in a followup
IMO. So do you agree with the change to add LIBXL_CFLAGS to virmocklibxl_la_CFLAGS?
Regards,
Jim