Re: [Xen-devel] [PATCH] VMX: fix vmx_{find,del}_msr() build

[Wei Liu <wei.liu2@xxxxxxxxxx>]

On Tue, Jul 17, 2018 at 09:39:03AM +0100, Andrew Cooper wrote:
> On 17/07/2018 07:57, Jan Beulich wrote:
> >>>> On 16.07.18 at 18:56, <andrew.cooper3@xxxxxxxxxx> wrote:
> >> On 16/07/18 17:46, Jan Beulich wrote:
> >>> For a reason that I can't explain, it is only the shim build that fails
> >>> for me with an older gcc due to the compiler not recognizing that
> >>> apparently uninitialized variables aren't really uninitialized.
> >> The only thing that comes to mind is some differences in CFLAGS et al. 
> >> There is nothing in kconfig which would plausibly impact that code.
> > Oh, right - looks like the shim build inherits the tool stack's CFLAGS,
> > which specify -O0 in debug builds. Otoh I'm surprised the hypervisor
> > builds at all with -O0, so I guess I'll have to look at that in some
> > more detail, not the least because xen/Rules.mk uses += to insert
> > its own -O<n>. In any event, comparing the two object files clearly
> > suggests a difference in optimization level.
> 
> Hmm - interesting.  I had intended not to inherit the tools dubious
> CFLAGS setting (nothing should use O0, not even for debugging), but how
> that I think back, that probably fell through the cracks in the
> associated chaos.
> 
> Nevertheless, all of our code should build at any optimisation level.  I
> know from other attempts that Xen also doesn't build at O3, and we could
> do with a way to select the optimisation level, independently of debug
> settings.

I agree with what you said but unfortunately reworking the build system
is rather low on my priority list.

Wei.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel