| Message ID | 1476110306-23754-1-git-send-email-eblake@redhat.com |
|---|---|
| State | New |
| Headers | show |
On 10 October 2016 at 15:38, Eric Blake <eblake@redhat.com> wrote: > C99 requires SIZE_MAX to be declared with the same type as the > integral promotion of size_t, but OSX mistakenly defines it as > an 'unsigned long long' expression even though size_t is only > 'unsigned long'. Rather than futzing around with whether size_t > is 32- or 64-bits wide (which would be needed if we cared about > using SIZE_T in a #if expression), let the compiler get the right > type for us by virtue of integer promotion. > > See also https://patchwork.ozlabs.org/patch/542327/ for an > instance where the wrong type trips us up if we don't fix it > for good in osdep.h. > > Some versions of glibc make a similar mistake with SSIZE_MAX; the > goal is that the approach of this patch could be copied to work > around that problem if it ever becomes important to us. > > Signed-off-by: Eric Blake <eblake@redhat.com> > > --- > v1 was here: > https://lists.gnu.org/archive/html/qemu-devel/2016-07/msg02520.html > > The topic recently came up again, and I noticed this patch sitting > on one of my older branches, so I've taken another shot at it. > https://lists.gnu.org/archive/html/qemu-devel/2016-10/msg00950.html > > v2: rewrite into a configure check (not sure if directly adding a > -D to QEMU_CFLAGS is the best, so advice welcome) Writing into config-host.mak would be preferable I think. > I lack easy access to a Mac box, so this is untested as to whether > it actually solves the issue... I ran a test with one of the cast workarounds locally removed, and this patch does allow the %zu version to build OK on OSX. > --- > include/qemu/osdep.h | 8 ++++++++ > configure | 13 +++++++++++++ > 2 files changed, 21 insertions(+) > > diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h > index 9e9fa61..22667f6 100644 > --- a/include/qemu/osdep.h > +++ b/include/qemu/osdep.h > @@ -141,6 +141,14 @@ extern int daemon(int, int); > # error Unknown pointer size > #endif > > +/* Mac OSX has a <stdint.h> bug that incorrectly defines SIZE_MAX with > + * the wrong type. Our replacement isn't usable in preprocessor > + * expressions, but it is sufficient for our needs. */ I guess if we get bitten by this it'll be obvious (break the OSX build) and we can fix it then. > +#ifdef SIZE_MAX_BROKEN > +#undef SIZE_MAX > +#define SIZE_MAX ((sizeof(char)) * -1) > +#endif > + > #ifndef MIN > #define MIN(a, b) (((a) < (b)) ? (a) : (b)) > #endif > diff --git a/configure b/configure > index 5751d8e..137ebb8 100755 > --- a/configure > +++ b/configure > @@ -1725,6 +1725,19 @@ if test "$cocoa" = "yes"; then > sdl=no > fi > > +# Some versions of Mac OS X incorrectly define SIZE_MAX > +cat > $TMPC << EOF > +#include <stdint.h> > +#include <stdio.h> > +int main(int argc, char *argv[]) { > + return printf("%zu", SIZE_MAX); > +} > +EOF > + > +if ! compile_object -Werror ; then > + QEMU_CFLAGS="-DSIZE_MAX_BROKEN $QEMU_CFLAGS" > +fi thanks -- PMM
On 10/10/2016 10:12 AM, Peter Maydell wrote: >> v2: rewrite into a configure check (not sure if directly adding a >> -D to QEMU_CFLAGS is the best, so advice welcome) > > Writing into config-host.mak would be preferable I think. > Okay, attempted in v4. >> I lack easy access to a Mac box, so this is untested as to whether >> it actually solves the issue... > > I ran a test with one of the cast workarounds locally removed, > and this patch does allow the %zu version to build OK on OSX. Thanks. But v3 leaves off Tested-by, because I still need to make sure my configure magic worked correctly :)
On 10/10/2016 01:38 PM, Eric Blake wrote: > On 10/10/2016 10:12 AM, Peter Maydell wrote: > >>> v2: rewrite into a configure check (not sure if directly adding a >>> -D to QEMU_CFLAGS is the best, so advice welcome) >> >> Writing into config-host.mak would be preferable I think. >> > > Okay, attempted in v4. I can't type today. I just sent v3, not v4. > >>> I lack easy access to a Mac box, so this is untested as to whether >>> it actually solves the issue... >> >> I ran a test with one of the cast workarounds locally removed, >> and this patch does allow the %zu version to build OK on OSX. > > Thanks. But v3 leaves off Tested-by, because I still need to make sure > my configure magic worked correctly :) >
diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h index 9e9fa61..22667f6 100644 --- a/include/qemu/osdep.h +++ b/include/qemu/osdep.h @@ -141,6 +141,14 @@ extern int daemon(int, int); # error Unknown pointer size #endif +/* Mac OSX has a <stdint.h> bug that incorrectly defines SIZE_MAX with + * the wrong type. Our replacement isn't usable in preprocessor + * expressions, but it is sufficient for our needs. */ +#ifdef SIZE_MAX_BROKEN +#undef SIZE_MAX +#define SIZE_MAX ((sizeof(char)) * -1) +#endif + #ifndef MIN #define MIN(a, b) (((a) < (b)) ? (a) : (b)) #endif diff --git a/configure b/configure index 5751d8e..137ebb8 100755 --- a/configure +++ b/configure @@ -1725,6 +1725,19 @@ if test "$cocoa" = "yes"; then sdl=no fi +# Some versions of Mac OS X incorrectly define SIZE_MAX +cat > $TMPC << EOF +#include <stdint.h> +#include <stdio.h> +int main(int argc, char *argv[]) { + return printf("%zu", SIZE_MAX); +} +EOF + +if ! compile_object -Werror ; then + QEMU_CFLAGS="-DSIZE_MAX_BROKEN $QEMU_CFLAGS" +fi + ########################################## # L2TPV3 probe
C99 requires SIZE_MAX to be declared with the same type as the integral promotion of size_t, but OSX mistakenly defines it as an 'unsigned long long' expression even though size_t is only 'unsigned long'. Rather than futzing around with whether size_t is 32- or 64-bits wide (which would be needed if we cared about using SIZE_T in a #if expression), let the compiler get the right type for us by virtue of integer promotion. See also https://patchwork.ozlabs.org/patch/542327/ for an instance where the wrong type trips us up if we don't fix it for good in osdep.h. Some versions of glibc make a similar mistake with SSIZE_MAX; the goal is that the approach of this patch could be copied to work around that problem if it ever becomes important to us. Signed-off-by: Eric Blake <eblake@redhat.com> --- v1 was here: https://lists.gnu.org/archive/html/qemu-devel/2016-07/msg02520.html The topic recently came up again, and I noticed this patch sitting on one of my older branches, so I've taken another shot at it. https://lists.gnu.org/archive/html/qemu-devel/2016-10/msg00950.html v2: rewrite into a configure check (not sure if directly adding a -D to QEMU_CFLAGS is the best, so advice welcome) I lack easy access to a Mac box, so this is untested as to whether it actually solves the issue... --- include/qemu/osdep.h | 8 ++++++++ configure | 13 +++++++++++++ 2 files changed, 21 insertions(+)