Patchwork [PATCHv2] configure: verify stdio.h

login
register
mail settings
Submitter Michael S. Tsirkin
Date Jan. 26, 2010, 4:01 p.m.
Message ID <20100126160147.GA3460@redhat.com>
Download mbox | patch
Permalink /patch/43715/
State New
Headers show

Comments

Michael S. Tsirkin - Jan. 26, 2010, 4:01 p.m.
Verify that stdio.h supports %lld %zd.
Some migw variants don't unless requested explicitly (see
http://www.mail-archive.com/mingw-w64-public@lists.sourceforge.net/msg00417.html)
), detect them early.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Juan Quintela <quintela@trasno.org>

---

changes from v1:
	removed leftover chunk

 configure |   20 ++++++++++++++++++++
 1 files changed, 20 insertions(+), 0 deletions(-)
Måns Rullgård - Jan. 26, 2010, 6:12 p.m.
"Michael S. Tsirkin" <mst@redhat.com> writes:

> Verify that stdio.h supports %lld %zd.
> Some migw variants don't unless requested explicitly (see
> http://www.mail-archive.com/mingw-w64-public@lists.sourceforge.net/msg00417.html)
> ), detect them early.
>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> Acked-by: Juan Quintela <quintela@trasno.org>
>
> ---
>
> changes from v1:
> 	removed leftover chunk
>
>  configure |   20 ++++++++++++++++++++
>  1 files changed, 20 insertions(+), 0 deletions(-)
>
> diff --git a/configure b/configure
> index 5631bbb..6ba06d6 100755
> --- a/configure
> +++ b/configure
> @@ -123,6 +123,26 @@ else
>      exit 1
>  fi
>
> +# Check that stdio.h compiler is sane: some
> +# mingw variants do not support %z %l that we rely on
> +cat > $TMPC <<EOF
> +#include <stddef.h>
> +#include <stdio.h>
> +size_t z = 1;
> +long long ll = 2;
> +int main(void) {
> +  printf("z=%zd;ll=%lld;\n", z, ll);
> +  return 0;
> +}
> +EOF
> +
> +if compile_prog && ($TMPE | grep "z=1;ll=2;" > /dev/null); then
> +  : C compiler works ok
> +else
> +    echo "ERROR: \"$cc\" does not have a working stdio.h"
> +    exit 1
> +fi
> +
>  check_define() {
>  cat > $TMPC <<EOF
>  #if !defined($1)

This makes cross-compiling impossible.  Is that really desirable?
Blue Swirl - Jan. 26, 2010, 7:31 p.m.
2010/1/26 Måns Rullgård <mans@mansr.com>:
> "Michael S. Tsirkin" <mst@redhat.com> writes:
>
>> Verify that stdio.h supports %lld %zd.
>> Some migw variants don't unless requested explicitly (see
>> http://www.mail-archive.com/mingw-w64-public@lists.sourceforge.net/msg00417.html)
>> ), detect them early.
>>
>> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
>> Acked-by: Juan Quintela <quintela@trasno.org>
>>
>> ---
>>
>> changes from v1:
>>       removed leftover chunk
>>
>>  configure |   20 ++++++++++++++++++++
>>  1 files changed, 20 insertions(+), 0 deletions(-)
>>
>> diff --git a/configure b/configure
>> index 5631bbb..6ba06d6 100755
>> --- a/configure
>> +++ b/configure
>> @@ -123,6 +123,26 @@ else
>>      exit 1
>>  fi
>>
>> +# Check that stdio.h compiler is sane: some
>> +# mingw variants do not support %z %l that we rely on
>> +cat > $TMPC <<EOF
>> +#include <stddef.h>
>> +#include <stdio.h>
>> +size_t z = 1;
>> +long long ll = 2;
>> +int main(void) {
>> +  printf("z=%zd;ll=%lld;\n", z, ll);
>> +  return 0;
>> +}
>> +EOF
>> +
>> +if compile_prog && ($TMPE | grep "z=1;ll=2;" > /dev/null); then
>> +  : C compiler works ok
>> +else
>> +    echo "ERROR: \"$cc\" does not have a working stdio.h"
>> +    exit 1
>> +fi
>> +
>>  check_define() {
>>  cat > $TMPC <<EOF
>>  #if !defined($1)
>
> This makes cross-compiling impossible.  Is that really desirable?

I'd rather drop %ll and %z.
Stefan Weil - Jan. 26, 2010, 9:44 p.m.
Michael S. Tsirkin schrieb:
> Verify that stdio.h supports %lld %zd.
> Some migw variants don't unless requested explicitly (see

migw -> mingw

I don't know any ming32 variant which supports %lld, %zd.
There is a new mingw-w64 were people are addressing the
problem, but that variant is unsupported by qemu.


> http://www.mail-archive.com/mingw-w64-public@lists.sourceforge.net/msg00417.html)
> ), detect them early.
>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> Acked-by: Juan Quintela <quintela@trasno.org>
>
> ---
>
> changes from v1:
> removed leftover chunk
>
> configure | 20 ++++++++++++++++++++
> 1 files changed, 20 insertions(+), 0 deletions(-)
>
> diff --git a/configure b/configure
> index 5631bbb..6ba06d6 100755
> --- a/configure
> +++ b/configure
> @@ -123,6 +123,26 @@ else
> exit 1
> fi
>
> +# Check that stdio.h compiler is sane: some
> +# mingw variants do not support %z %l that we rely on
> +cat > $TMPC <<EOF
> +#include <stddef.h>
> +#include <stdio.h>
> +size_t z = 1;
> +long long ll = 2;
> +int main(void) {
> + printf("z=%zd;ll=%lld;\n", z, ll);
> + return 0;
> +}
> +EOF
> +
> +if compile_prog && ($TMPE | grep "z=1;ll=2;" > /dev/null); then
> + : C compiler works ok
> +else
> + echo "ERROR: \"$cc\" does not have a working stdio.h"
> + exit 1
> +fi
> +
> check_define() {
> cat > $TMPC <<EOF
> #if !defined($1)


Tests of %lld must use a 64 bit pattern (ll = 0x0123456789abcdefLL).
Otherwise, %lld, %ld and %d with ll = 1 will all print 1 on little endian
machines, and the %lld test won't detect missing support.

The problem is now several years old, it can be avoided for really
important output, so it seems acceptable to wait another period
of time until a working mingw is supported by qemu.

A test which prevents qemu builds on windows with current mingw32
or cross builds with current debian cross mingw32 would not help.
Both variants work for me even without full format support.

Regards,
Stefan Weil
Michael S. Tsirkin - Jan. 27, 2010, 9:13 a.m.
On Tue, Jan 26, 2010 at 10:44:44PM +0100, Stefan Weil wrote:
> Michael S. Tsirkin schrieb:
> > Verify that stdio.h supports %lld %zd.
> > Some migw variants don't unless requested explicitly (see
> 
> migw -> mingw
> 
> I don't know any ming32 variant which supports %lld, %zd.
> There is a new mingw-w64 were people are addressing the
> problem, 

Did you try with -D__USE_MINGW_ANSI_STDIO=1 ?
This is what the link below mentions.

> but that variant is unsupported by qemu.

What are the issues?

> 
> > http://www.mail-archive.com/mingw-w64-public@lists.sourceforge.net/msg00417.html)
> > ), detect them early.
> >
> > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > Acked-by: Juan Quintela <quintela@trasno.org>
> >
> > ---
> >
> > changes from v1:
> > removed leftover chunk
> >
> > configure | 20 ++++++++++++++++++++
> > 1 files changed, 20 insertions(+), 0 deletions(-)
> >
> > diff --git a/configure b/configure
> > index 5631bbb..6ba06d6 100755
> > --- a/configure
> > +++ b/configure
> > @@ -123,6 +123,26 @@ else
> > exit 1
> > fi
> >
> > +# Check that stdio.h compiler is sane: some
> > +# mingw variants do not support %z %l that we rely on
> > +cat > $TMPC <<EOF
> > +#include <stddef.h>
> > +#include <stdio.h>
> > +size_t z = 1;
> > +long long ll = 2;
> > +int main(void) {
> > + printf("z=%zd;ll=%lld;\n", z, ll);
> > + return 0;
> > +}
> > +EOF
> > +
> > +if compile_prog && ($TMPE | grep "z=1;ll=2;" > /dev/null); then
> > + : C compiler works ok
> > +else
> > + echo "ERROR: \"$cc\" does not have a working stdio.h"
> > + exit 1
> > +fi
> > +
> > check_define() {
> > cat > $TMPC <<EOF
> > #if !defined($1)
> 
> 
> Tests of %lld must use a 64 bit pattern (ll = 0x0123456789abcdefLL).
> Otherwise, %lld, %ld and %d with ll = 1 will all print 1 on little endian
> machines, and the %lld test won't detect missing support.
> 
> The problem is now several years old, it can be avoided for really
> important output, so it seems acceptable to wait another period
> of time until a working mingw is supported by qemu.
> 
> A test which prevents qemu builds on windows with current mingw32
> or cross builds with current debian cross mingw32 would not help.
> Both variants work for me even without full format support.
> 
> Regards,
> Stefan Weil
Stefan Weil - Jan. 27, 2010, 7:02 p.m.
Michael S. Tsirkin schrieb:
> On Tue, Jan 26, 2010 at 10:44:44PM +0100, Stefan Weil wrote:
>> Michael S. Tsirkin schrieb:
>>> Verify that stdio.h supports %lld %zd.
>>> Some migw variants don't unless requested explicitly (see
>> migw -> mingw
>>
>> I don't know any ming32 variant which supports %lld, %zd.
>> There is a new mingw-w64 were people are addressing the
>> problem,
>
> Did you try with -D__USE_MINGW_ANSI_STDIO=1 ?
> This is what the link below mentions.


New information on this issue:

mingw-runtime 3.15 (2008-09-07) added support for
standard format conversion:

http://sourceforge.net/project/shownotes.php?release_id=24832

The release note also says how to select MS or C99/POSIX
format strings.

Debian still does not include that version:

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=498529

Maybe someone knows how we can help to get the new release
for debian. It's needed for cross compilation (at least for
users who don't want to install it manually).

>
>> but that variant is unsupported by qemu.
>
> What are the issues?
>

Good question. I don't know because I did not try it.
It was only a guess that qemu would not work with
mingw-w64 without modifications.
Michael S. Tsirkin - Jan. 27, 2010, 7:12 p.m.
On Wed, Jan 27, 2010 at 08:02:26PM +0100, Stefan Weil wrote:
> Michael S. Tsirkin schrieb:
> > On Tue, Jan 26, 2010 at 10:44:44PM +0100, Stefan Weil wrote:
> >> Michael S. Tsirkin schrieb:
> >>> Verify that stdio.h supports %lld %zd.
> >>> Some migw variants don't unless requested explicitly (see
> >> migw -> mingw
> >>
> >> I don't know any ming32 variant which supports %lld, %zd.
> >> There is a new mingw-w64 were people are addressing the
> >> problem,
> >
> > Did you try with -D__USE_MINGW_ANSI_STDIO=1 ?
> > This is what the link below mentions.
> 
> 
> New information on this issue:
> 
> mingw-runtime 3.15 (2008-09-07) added support for
> standard format conversion:
> 
> http://sourceforge.net/project/shownotes.php?release_id=24832
> 
> The release note also says how to select MS or C99/POSIX
> format strings.
> 
> Debian still does not include that version:
> 
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=498529
> 
> Maybe someone knows how we can help to get the new release
> for debian. It's needed for cross compilation (at least for
> users who don't want to install it manually).

Heh, configure script runs the program it's built
in a couple of places. This probably does not
work for cross-builds:

	if compile_prog "" "" ; then
	$TMPE && bigendian="yes"
	else
	echo big/little test failed
	fi

likely works by luck for build to intel, because error
is interpreted as little-endian.

> >
> >> but that variant is unsupported by qemu.
> >
> > What are the issues?
> >
> 
> Good question. I don't know because I did not try it.
> It was only a guess that qemu would not work with
> mingw-w64 without modifications.
Måns Rullgård - Jan. 27, 2010, 9:48 p.m.
"Michael S. Tsirkin" <mst@redhat.com> writes:

> Heh, configure script runs the program it's built
> in a couple of places. This probably does not
> work for cross-builds:
>
> 	if compile_prog "" "" ; then
> 	$TMPE && bigendian="yes"
> 	else
> 	echo big/little test failed
> 	fi
>
> likely works by luck for build to intel, because error
> is interpreted as little-endian.

In FFmpeg we check the endianness like this:

check_cc <<EOF || die "endian test failed"
unsigned int endian = 'B' << 24 | 'I' << 16 | 'G' << 8 | 'E';
EOF
od -t x1 $TMPO | grep -q '42 *49 *47 *45' && enable bigendian

This works on every combination of build and target system we use.
Michael S. Tsirkin - Jan. 28, 2010, 5:48 p.m.
On Wed, Jan 27, 2010 at 08:02:26PM +0100, Stefan Weil wrote:
> Michael S. Tsirkin schrieb:
> > On Tue, Jan 26, 2010 at 10:44:44PM +0100, Stefan Weil wrote:
> >> Michael S. Tsirkin schrieb:
> >>> Verify that stdio.h supports %lld %zd.
> >>> Some migw variants don't unless requested explicitly (see
> >> migw -> mingw
> >>
> >> I don't know any ming32 variant which supports %lld, %zd.
> >> There is a new mingw-w64 were people are addressing the
> >> problem,
> >
> > Did you try with -D__USE_MINGW_ANSI_STDIO=1 ?
> > This is what the link below mentions.
> 
> 
> New information on this issue:
> 
> mingw-runtime 3.15 (2008-09-07) added support for
> standard format conversion:
> 
> http://sourceforge.net/project/shownotes.php?release_id=24832
> 
> The release note also says how to select MS or C99/POSIX
> format strings.
> 
> Debian still does not include that version:
> 
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=498529
> 
> Maybe someone knows how we can help to get the new release
> for debian. It's needed for cross compilation (at least for
> users who don't want to install it manually).

Format disk, install Fedora :)

> >
> >> but that variant is unsupported by qemu.
> >
> > What are the issues?
> >
> 
> Good question. I don't know because I did not try it.
> It was only a guess that qemu would not work with
> mingw-w64 without modifications.
> 
>

Patch

diff --git a/configure b/configure
index 5631bbb..6ba06d6 100755
--- a/configure
+++ b/configure
@@ -123,6 +123,26 @@  else
     exit 1
 fi
 
+# Check that stdio.h compiler is sane: some
+# mingw variants do not support %z %l that we rely on
+cat > $TMPC <<EOF
+#include <stddef.h>
+#include <stdio.h>
+size_t z = 1;
+long long ll = 2;
+int main(void) {
+  printf("z=%zd;ll=%lld;\n", z, ll);
+  return 0;
+}
+EOF
+
+if compile_prog && ($TMPE | grep "z=1;ll=2;" > /dev/null); then
+  : C compiler works ok
+else
+    echo "ERROR: \"$cc\" does not have a working stdio.h"
+    exit 1
+fi
+
 check_define() {
 cat > $TMPC <<EOF
 #if !defined($1)