On czw, 2014-04-10 at 04:01 +0100, Al Viro wrote: > On Wed, Apr 09, 2014 at 10:28:59PM +0200, Lars-Peter Clausen wrote: > > > I had a look at the sparse code and the problem is that it sets the > > default return type of sizeof according to the type of the host it > > was compiled on (either unsigned int or unsigned long). It can be > > overwritten by switches like -m32, but of course wont work when > > cross compiling. So if your host system is 64bit, but your target > > system is 32bit you'll get that warning. > > Not that simple. First of all, you *need* to tell sparse what target > to expect; size_t is the least of your troubles - sizeof(long) has > far more widespread impact. If you don't get -m64 or -m32 in CFLAGS > (and on quite a few cross-builds you get it, simply because the target > is biarch and gcc itself needs to know what to generate), you need > to set it in CHECKFLAGS, along with other target-specific things. > Example: > arch/ia64/Makefile:21:CHECKFLAGS += -m64 -D__ia64=1 -D__ia64__=1 > -D_LP64 -D__LP64__ > > With the defaults being what they are (since commit 7aa79f "Adding default for > m64/m32 handle"), we probably need explicit -m32 in CHECKFLAGS for a bunch > of 32bit targets. Defaults are iffy, BTW - it's not even "do as host does", > it's "64bit if the host is amd64, 32bit otherwise" ;-/ > > Again, CHECKFLAGS need to be set; that's normally done in arch/*/Makefile.
In my case it was cross compile on x86_64 for ARM so adding -m32 to CHECKFLAGS in arch/arm/Makefile helped. Best regards, Krzysztof -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
