Re: [Xen-devel] [PATCH 2/4] Adding support for coverage information

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 04.02.13 at 16:29, Frediano Ziglio <frediano.ziglio@xxxxxxxxxx> wrote:
> On Mon, 2013-02-04 at 09:39 +0000, Jan Beulich wrote:
>> >>> On 01.02.13 at 15:48, Frediano Ziglio <frediano.ziglio@xxxxxxxxxx> wrote:
>> > +ifeq ($(coverage),y)
>> > +$(filter-out %.init.o,$(obj-y) $(obj-bin-y)): CFLAGS += -fprofile-arcs 
>> > -ftest-coverage -DTEST_COVERAGE
>> > +endif
>> 
>> For one - isn't simply using $(obj-y) here sufficient (i.e. without the
>> $(filter-out ...))?
>> 
>> And second, I would thing this then ought to become a single line:
>> 
>> $(obj-$(coverage)): CFLAGS += -fprofile-arcs -ftest-coverage -DTEST_COVERAGE
>> 
> 
> Yes, it works fine and seems to compile all proper files but I don't
> understand why.
> What's the difference between $(obj-bin-y) and $(obj-y) ??

The former is the set of all .init.o files, which (when building with
clang instead of gcc) must not undergo certain transformations
(and aren't worth it, as init-time only code isn't really performance
critical).

>> > +
>> > +void init_coverage(void)
>> 
>> __init
>> 
> 
> I removed from previous and put in the header as suggestion.

Not sure what you mean here, but just in case - __init
annotations do specifically not belong on declarations, but
only on definitions.

>> > +#if BITS_PER_LONG >= 64
>> > +typedef long gcov_type;
>> > +#else
>> > +typedef long long gcov_type;
>> > +#endif
>> 
>> What's this???
>> 
> 
> This came from Linux, probably for Xen a simple
> 
> typedef long gcov_type
> 
> is ok.

No. Your goal - just to reiterate - must be not to use types the
size of which depends on the guest word size.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel