| Message ID | 1447414307-5295-1-git-send-email-berrange@redhat.com |
|---|---|
| State | New |
| Headers | show |
* Daniel P. Berrange (berrange@redhat.com) wrote: > Developers on 64-bit machines will often try to perform a > 32-bit build of QEMU by running > > ./configure --extra-cflags="-m32" > > Unfortunately if PKG_CONFIG_LIBDIR is not set to point to > the location of the 32-bit pkg-config files, then configure > will silently pick up the 64-bit pkg-config files and still > succeed. Having fallen into this trap, thanks; seems to do the trick on RHEL7.x: # /root/try/../qemu/configure --prefix=/root/try --enable-trace-backends=stderr --extra-cflags="-m32" --extra-ldflags="-m32" ERROR: sizeof(size_t) doesn't match GLIB_SIZEOF_SIZE_T. You probably need to set PKG_CONFIG_LIBDIR to point to the right pkg-config files for your build target # export PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig # /root/try/../qemu/configure --prefix=/root/try --enable-trace-backends=stderr --extra-cflags="-m32" --extra-ldflags="-m32" Install prefix /root/try BIOS directory /root/try/share/qemu .... and the gsize def ( https://developer.gnome.org/glib/stable/glib-Basic-Types.html#gsize ) does say it should always match size_t size. Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com> Dave > > This causes a problem for glib because it means QEMU will > be pulling in /usr/lib64/glib-2.0/include/glibconfig.h > instead of /usr/lib/glib-2.0/include/glibconfig.h > > This causes problems because the 'gsize' type (defined as > 'unsigned long') will no longer be fully compatible with > the 'size_t' type (defined as 'unsigned int'). Although > both are the same size, the compiler refuses to allow > casts from 'unsigned long *' to 'unsigned int *' as they > are different pointer types. This results in non-obvious > compiler errors when building QEMU eg > > qga/commands-posix.c: In function ‘qmp_guest_set_user_password’: > qga/commands-posix.c:1912:55: error: passing argument 2 of ‘g_base64_decode’ from incompatible pointer type [-Werror=incompatible-pointer-types] > rawpasswddata = (char *)g_base64_decode(password, &rawpasswdlen); > ^ > In file included from /usr/include/glib-2.0/glib.h:35:0, > from qga/commands-posix.c:14: > /usr/include/glib-2.0/glib/gbase64.h:52:9: note: expected ‘gsize * {aka long unsigned int *}’ but argument is of type ‘size_t * {aka unsigned int *}’ > guchar *g_base64_decode (const gchar *text, > ^ > cc1: all warnings being treated as errors > > To detect this problem, add a check to configure that > verifies that GLIB_SIZEOF_SIZE_T matches sizeof(size_t). > If this fails print a warning suggesting that the dev > probably needs to set PKG_CONFIG_LIBDIR. > > On Fedora x86_64 it passes with any of: > > # ./configure > # PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig ./configure --extra-cflags="-m32" > # PKG_CONFIG_LIBDIR=/usr/lib64/pkgconfig ./configure --extra-cflags="-m64" > > And fails with a mis-match > > # PKG_CONFIG_LIBDIR=/usr/lib64/pkgconfig ./configure --extra-cflags="-m32" > # PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig ./configure --extra-cflags="-m64" > > ERROR: sizeof(size_t) doesn't match GLIB_SIZEOF_SIZE_T. > You probably need to set PKG_CONFIG_LIBDIR > to point to the right pkg-config files for your > build target > > Signed-off-by: Daniel P. Berrange <berrange@redhat.com> > --- > Changed in v2: > > - Add 'return 0' to silence compiler warning on some > platforms about returning without a value > > configure | 21 +++++++++++++++++++++ > 1 file changed, 21 insertions(+) > > diff --git a/configure b/configure > index f75df4b..c53334f 100755 > --- a/configure > +++ b/configure > @@ -2939,6 +2939,27 @@ for i in $glib_modules; do > fi > done > > +# Sanity check that the current size_t matches the > +# size that glib thinks it should be. This catches > +# problems on multi-arch where people try to build > +# 32-bit QEMU while pointing at 64-bit glib headers > +cat > $TMPC <<EOF > +#include <glib.h> > +#include <unistd.h> > + > +int main(void) { > + G_STATIC_ASSERT(sizeof(size_t) == GLIB_SIZEOF_SIZE_T); > + return 0; > +} > +EOF > + > +if ! compile_prog "-Werror $CFLAGS" "$LIBS" ; then > + error_exit "sizeof(size_t) doesn't match GLIB_SIZEOF_SIZE_T."\ > + "You probably need to set PKG_CONFIG_LIBDIR"\ > + "to point to the right pkg-config files for your"\ > + "build target" > +fi > + > # g_test_trap_subprocess added in 2.38. Used by some tests. > glib_subprocess=yes > if ! $pkg_config --atleast-version=2.38 glib-2.0; then > -- > 2.5.0 > -- Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
On Fr, 2015-11-13 at 11:31 +0000, Daniel P. Berrange wrote: > Developers on 64-bit machines will often try to perform a > 32-bit build of QEMU by running > > ./configure --extra-cflags="-m32" > > Unfortunately if PKG_CONFIG_LIBDIR is not set to point to > the location of the 32-bit pkg-config files, then configure > will silently pick up the 64-bit pkg-config files and still > succeed. While being at it: I think it would be a good idea to record PKG_CONFIG_LIBDIR in config.status, otherwise it'll be lost the next time configure is touched and make figures it needs to re-run configure.
On 13 November 2015 at 11:31, Daniel P. Berrange <berrange@redhat.com> wrote: > Developers on 64-bit machines will often try to perform a > 32-bit build of QEMU by running > > ./configure --extra-cflags="-m32" > > Unfortunately if PKG_CONFIG_LIBDIR is not set to point to > the location of the 32-bit pkg-config files, then configure > will silently pick up the 64-bit pkg-config files and still > succeed. Will this also handle all the other -config programs that seem to have sprouted up? Looking in configure we also call sdl-config sdl2-config libgcrypt-config aalib-config curl-config thanks -- PMM
On Fri, Nov 13, 2015 at 01:01:52PM +0000, Peter Maydell wrote: > On 13 November 2015 at 11:31, Daniel P. Berrange <berrange@redhat.com> wrote: > > Developers on 64-bit machines will often try to perform a > > 32-bit build of QEMU by running > > > > ./configure --extra-cflags="-m32" > > > > Unfortunately if PKG_CONFIG_LIBDIR is not set to point to > > the location of the 32-bit pkg-config files, then configure > > will silently pick up the 64-bit pkg-config files and still > > succeed. > > Will this also handle all the other -config programs that > seem to have sprouted up? Looking in configure we also call > sdl-config > sdl2-config > libgcrypt-config > aalib-config > curl-config For custom non-pkg-config programs there's not any good general purpose solution for them to look for 32-bit versions while on 64-bit hosts, at least not when 32-bit and 64-bit are both installed into the same /usr prefix. If the 32-bit stuff is installed in a completely separate root prefix, then you can sometimes update $PATH to ensure the right -config binary is found. This is a good reason why libs are encouraged to switch to using pkg-config instead of home-grown alternatives Regards, Daniel
On Fri, Nov 13, 2015 at 02:01:12PM +0100, Gerd Hoffmann wrote: > On Fr, 2015-11-13 at 11:31 +0000, Daniel P. Berrange wrote: > > Developers on 64-bit machines will often try to perform a > > 32-bit build of QEMU by running > > > > ./configure --extra-cflags="-m32" > > > > Unfortunately if PKG_CONFIG_LIBDIR is not set to point to > > the location of the 32-bit pkg-config files, then configure > > will silently pick up the 64-bit pkg-config files and still > > succeed. > > While being at it: I think it would be a good idea to record > PKG_CONFIG_LIBDIR in config.status, otherwise it'll be lost the next > time configure is touched and make figures it needs to re-run configure. In the spirit of doing one logical change per patch, I'm sending a separate patch to cover this suggestion. Regards, Daniel
On 13 November 2015 at 11:31, Daniel P. Berrange <berrange@redhat.com> wrote: > Developers on 64-bit machines will often try to perform a > 32-bit build of QEMU by running > > ./configure --extra-cflags="-m32" > > Unfortunately if PKG_CONFIG_LIBDIR is not set to point to > the location of the 32-bit pkg-config files, then configure > will silently pick up the 64-bit pkg-config files and still > succeed. > > This causes a problem for glib because it means QEMU will > be pulling in /usr/lib64/glib-2.0/include/glibconfig.h > instead of /usr/lib/glib-2.0/include/glibconfig.h > > This causes problems because the 'gsize' type (defined as > 'unsigned long') will no longer be fully compatible with > the 'size_t' type (defined as 'unsigned int'). Although > both are the same size, the compiler refuses to allow > casts from 'unsigned long *' to 'unsigned int *' as they > are different pointer types. This results in non-obvious > compiler errors when building QEMU eg > > qga/commands-posix.c: In function ‘qmp_guest_set_user_password’: > qga/commands-posix.c:1912:55: error: passing argument 2 of ‘g_base64_decode’ from incompatible pointer type [-Werror=incompatible-pointer-types] > rawpasswddata = (char *)g_base64_decode(password, &rawpasswdlen); > ^ > In file included from /usr/include/glib-2.0/glib.h:35:0, > from qga/commands-posix.c:14: > /usr/include/glib-2.0/glib/gbase64.h:52:9: note: expected ‘gsize * {aka long unsigned int *}’ but argument is of type ‘size_t * {aka unsigned int *}’ > guchar *g_base64_decode (const gchar *text, > ^ > cc1: all warnings being treated as errors > > To detect this problem, add a check to configure that > verifies that GLIB_SIZEOF_SIZE_T matches sizeof(size_t). > If this fails print a warning suggesting that the dev > probably needs to set PKG_CONFIG_LIBDIR. > > On Fedora x86_64 it passes with any of: > > # ./configure > # PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig ./configure --extra-cflags="-m32" > # PKG_CONFIG_LIBDIR=/usr/lib64/pkgconfig ./configure --extra-cflags="-m64" > > And fails with a mis-match > > # PKG_CONFIG_LIBDIR=/usr/lib64/pkgconfig ./configure --extra-cflags="-m32" > # PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig ./configure --extra-cflags="-m64" > > ERROR: sizeof(size_t) doesn't match GLIB_SIZEOF_SIZE_T. > You probably need to set PKG_CONFIG_LIBDIR > to point to the right pkg-config files for your > build target > > Signed-off-by: Daniel P. Berrange <berrange@redhat.com> I just noticed this never got applied to master -- did it just fall through the cracks, or is there an issue with it? thanks -- PMM
On Mon, Jan 25, 2016 at 02:42:10PM +0000, Peter Maydell wrote: > On 13 November 2015 at 11:31, Daniel P. Berrange <berrange@redhat.com> wrote: > > Developers on 64-bit machines will often try to perform a > > 32-bit build of QEMU by running > > > > ./configure --extra-cflags="-m32" > > > > Unfortunately if PKG_CONFIG_LIBDIR is not set to point to > > the location of the 32-bit pkg-config files, then configure > > will silently pick up the 64-bit pkg-config files and still > > succeed. > > > > This causes a problem for glib because it means QEMU will > > be pulling in /usr/lib64/glib-2.0/include/glibconfig.h > > instead of /usr/lib/glib-2.0/include/glibconfig.h > > > > This causes problems because the 'gsize' type (defined as > > 'unsigned long') will no longer be fully compatible with > > the 'size_t' type (defined as 'unsigned int'). Although > > both are the same size, the compiler refuses to allow > > casts from 'unsigned long *' to 'unsigned int *' as they > > are different pointer types. This results in non-obvious > > compiler errors when building QEMU eg > > > > qga/commands-posix.c: In function ‘qmp_guest_set_user_password’: > > qga/commands-posix.c:1912:55: error: passing argument 2 of ‘g_base64_decode’ from incompatible pointer type [-Werror=incompatible-pointer-types] > > rawpasswddata = (char *)g_base64_decode(password, &rawpasswdlen); > > ^ > > In file included from /usr/include/glib-2.0/glib.h:35:0, > > from qga/commands-posix.c:14: > > /usr/include/glib-2.0/glib/gbase64.h:52:9: note: expected ‘gsize * {aka long unsigned int *}’ but argument is of type ‘size_t * {aka unsigned int *}’ > > guchar *g_base64_decode (const gchar *text, > > ^ > > cc1: all warnings being treated as errors > > > > To detect this problem, add a check to configure that > > verifies that GLIB_SIZEOF_SIZE_T matches sizeof(size_t). > > If this fails print a warning suggesting that the dev > > probably needs to set PKG_CONFIG_LIBDIR. > > > > On Fedora x86_64 it passes with any of: > > > > # ./configure > > # PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig ./configure --extra-cflags="-m32" > > # PKG_CONFIG_LIBDIR=/usr/lib64/pkgconfig ./configure --extra-cflags="-m64" > > > > And fails with a mis-match > > > > # PKG_CONFIG_LIBDIR=/usr/lib64/pkgconfig ./configure --extra-cflags="-m32" > > # PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig ./configure --extra-cflags="-m64" > > > > ERROR: sizeof(size_t) doesn't match GLIB_SIZEOF_SIZE_T. > > You probably need to set PKG_CONFIG_LIBDIR > > to point to the right pkg-config files for your > > build target > > > > Signed-off-by: Daniel P. Berrange <berrange@redhat.com> > > I just noticed this never got applied to master -- did it just fall through > the cracks, or is there an issue with it? AFAIK, no one raised a blocking issue with it. Regards, Daniel
On 25 January 2016 at 14:47, Daniel P. Berrange <berrange@redhat.com> wrote:
> AFAIK, no one raised a blocking issue with it.
I've just found one :-) which is that it fails on my OSX build system:
config-temp/qemu-conf.c:5:4: error: unused typedef '_GStaticAssert_5'
[-Werror,-Wunused-local-typedef]
G_STATIC_ASSERT(sizeof(size_t) == GLIB_SIZEOF_SIZE_T);
^
/sw/include/glib-2.0/glib/gmacros.h:143:96: note: expanded from macro
'G_STATIC_ASSERT'
#define G_STATIC_ASSERT(expr) typedef struct { char
Compile_Time_Assertion[(expr) ? 1 : -1]; } G_PASTE (_GStaticAssert_,
__LINE__)
^
/sw/include/glib-2.0/glib/gmacros.h:142:47: note: expanded from macro 'G_PASTE'
#define G_PASTE(identifier1,identifier2) G_PASTE_ARGS
(identifier1, identifier2)
^
/sw/include/glib-2.0/glib/gmacros.h:141:47: note: expanded from macro
'G_PASTE_ARGS'
#define G_PASTE_ARGS(identifier1,identifier2) identifier1 ## identifier2
^
<scratch space>:170:1: note: expanded from here
_GStaticAssert_5
^
1 error generated.
The annotation of gmacros.h to mark the typedef as deliberately unused
didn't happen until after the version of glib I have. G_STATIC_ASSERT
is probably best avoided in QEMU code (we have QEMU_BUILD_BUG_ON
anyway).
thanks
-- PMM
On Mon, Jan 25, 2016 at 02:58:22PM +0000, Peter Maydell wrote: > On 25 January 2016 at 14:47, Daniel P. Berrange <berrange@redhat.com> wrote: > > AFAIK, no one raised a blocking issue with it. > > I've just found one :-) which is that it fails on my OSX build system: > > config-temp/qemu-conf.c:5:4: error: unused typedef '_GStaticAssert_5' > [-Werror,-Wunused-local-typedef] > G_STATIC_ASSERT(sizeof(size_t) == GLIB_SIZEOF_SIZE_T); > ^ > /sw/include/glib-2.0/glib/gmacros.h:143:96: note: expanded from macro > 'G_STATIC_ASSERT' > #define G_STATIC_ASSERT(expr) typedef struct { char > Compile_Time_Assertion[(expr) ? 1 : -1]; } G_PASTE (_GStaticAssert_, > __LINE__) > > ^ > /sw/include/glib-2.0/glib/gmacros.h:142:47: note: expanded from macro 'G_PASTE' > #define G_PASTE(identifier1,identifier2) G_PASTE_ARGS > (identifier1, identifier2) > ^ > /sw/include/glib-2.0/glib/gmacros.h:141:47: note: expanded from macro > 'G_PASTE_ARGS' > #define G_PASTE_ARGS(identifier1,identifier2) identifier1 ## identifier2 > ^ > <scratch space>:170:1: note: expanded from here > _GStaticAssert_5 > ^ > 1 error generated. > > The annotation of gmacros.h to mark the typedef as deliberately unused > didn't happen until after the version of glib I have. G_STATIC_ASSERT > is probably best avoided in QEMU code (we have QEMU_BUILD_BUG_ON > anyway). I'm travelling for a few days, but will investigate and re-post an update once i return. Regards, Daniel
diff --git a/configure b/configure index f75df4b..c53334f 100755 --- a/configure +++ b/configure @@ -2939,6 +2939,27 @@ for i in $glib_modules; do fi done +# Sanity check that the current size_t matches the +# size that glib thinks it should be. This catches +# problems on multi-arch where people try to build +# 32-bit QEMU while pointing at 64-bit glib headers +cat > $TMPC <<EOF +#include <glib.h> +#include <unistd.h> + +int main(void) { + G_STATIC_ASSERT(sizeof(size_t) == GLIB_SIZEOF_SIZE_T); + return 0; +} +EOF + +if ! compile_prog "-Werror $CFLAGS" "$LIBS" ; then + error_exit "sizeof(size_t) doesn't match GLIB_SIZEOF_SIZE_T."\ + "You probably need to set PKG_CONFIG_LIBDIR"\ + "to point to the right pkg-config files for your"\ + "build target" +fi + # g_test_trap_subprocess added in 2.38. Used by some tests. glib_subprocess=yes if ! $pkg_config --atleast-version=2.38 glib-2.0; then
Developers on 64-bit machines will often try to perform a 32-bit build of QEMU by running ./configure --extra-cflags="-m32" Unfortunately if PKG_CONFIG_LIBDIR is not set to point to the location of the 32-bit pkg-config files, then configure will silently pick up the 64-bit pkg-config files and still succeed. This causes a problem for glib because it means QEMU will be pulling in /usr/lib64/glib-2.0/include/glibconfig.h instead of /usr/lib/glib-2.0/include/glibconfig.h This causes problems because the 'gsize' type (defined as 'unsigned long') will no longer be fully compatible with the 'size_t' type (defined as 'unsigned int'). Although both are the same size, the compiler refuses to allow casts from 'unsigned long *' to 'unsigned int *' as they are different pointer types. This results in non-obvious compiler errors when building QEMU eg qga/commands-posix.c: In function ‘qmp_guest_set_user_password’: qga/commands-posix.c:1912:55: error: passing argument 2 of ‘g_base64_decode’ from incompatible pointer type [-Werror=incompatible-pointer-types] rawpasswddata = (char *)g_base64_decode(password, &rawpasswdlen); ^ In file included from /usr/include/glib-2.0/glib.h:35:0, from qga/commands-posix.c:14: /usr/include/glib-2.0/glib/gbase64.h:52:9: note: expected ‘gsize * {aka long unsigned int *}’ but argument is of type ‘size_t * {aka unsigned int *}’ guchar *g_base64_decode (const gchar *text, ^ cc1: all warnings being treated as errors To detect this problem, add a check to configure that verifies that GLIB_SIZEOF_SIZE_T matches sizeof(size_t). If this fails print a warning suggesting that the dev probably needs to set PKG_CONFIG_LIBDIR. On Fedora x86_64 it passes with any of: # ./configure # PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig ./configure --extra-cflags="-m32" # PKG_CONFIG_LIBDIR=/usr/lib64/pkgconfig ./configure --extra-cflags="-m64" And fails with a mis-match # PKG_CONFIG_LIBDIR=/usr/lib64/pkgconfig ./configure --extra-cflags="-m32" # PKG_CONFIG_LIBDIR=/usr/lib/pkgconfig ./configure --extra-cflags="-m64" ERROR: sizeof(size_t) doesn't match GLIB_SIZEOF_SIZE_T. You probably need to set PKG_CONFIG_LIBDIR to point to the right pkg-config files for your build target Signed-off-by: Daniel P. Berrange <berrange@redhat.com> --- Changed in v2: - Add 'return 0' to silence compiler warning on some platforms about returning without a value configure | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+)