On Mon, 2015-02-16 at 10:54 -0500, Boris Ostrovsky wrote: > I am not sure what the standard practice is for LIBXL_HAVE_ macros. I > see a couple of examples where '#ifdef LIBXL_HAVE_*" is used in libxl > code, which is why I have it here as well. > Yes, that is the case for two of them, AFAICS: * LIBXL_HAVE_NO_SUSPEND_RESUME * LIBXL_HAVE_PSR_CMT If you look at how these two are defined, you'll see that the definition is actually conditional: #if defined(__arm__) || defined(__aarch64__) #define LIBXL_HAVE_NO_SUSPEND_RESUME 1 #endif #if defined(__i386__) || defined(__x86_64__) #define LIBXL_HAVE_PSR_CMT 1 #endif Which is, IMO, why we need to take special care of them in libxl and xl code. All other macros, which are always and unconditionally #defined to 1, we just forget about them, and leave it to the caller. > This is probably question for maintainers. > Sure! :-) > > Also, ISTR that the first change that actually changes the API should > > bump the MAJOR in libxl's Makefile. I'm not sure this change qualifies > > as such, as you're adding stuff, not altering existing data structs > > (e.g., by adding fields, etc). Personally, I think it does, but I'm > > leaving this to tools maintainers. > > libxl.h seems to suggest that API version is changed only when we make > an incompatible change to the library. In my mind new interface is does > not break compatibility so I didn't think a bump would be necessary. > Yeah, that was why I was not sure. Re-thinking about it, you probably are right, no need for it for now. Regards, Dario
Attachment:
signature.asc
Description: This is a digitally signed message part