Skip to contentSkip to conversation

oslib-win32: only provide localtime_r/gmtime_r if missing

PATCH Superseded Daniel P. Berrange 7 participants 21 messages Jul 31, 2015 – Sep 10, 2015

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 oslib-win32: only provide localtime_r/gmtime_r if missing

The oslib-win32 file currently provides a localtime_r and gmtime_r replacement unconditionally. Some versions of Mingw64 would provide crude macros for localtime_r/gmtime_r which QEMU takes care to disable. Latest versions of Mingw64 now provide actual functions for localtime_r/gmtime_r, but with a twist that you have to include unistd.h or pthread.h before including time.h. By luck some files in QEMU have such an include order, resulting in compile errors:

CC util/osdep.o
In file included from include/qemu-common.h:48:0,

                 from util/osdep.c:48:
include/sysemu/os-win32.h:77:12: error: redundant redeclaration of 'gmtime_r' [-Werror=redundant-decls]
 struct tm *gmtime_r(const time_t *timep, struct tm *result);
            ^
In file included from include/qemu-common.h:35:0,
                 from util/osdep.c:48:
/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:272:107: note: previous definition of 'gmtime_r' was here
In file included from include/qemu-common.h:48:0,
                 from util/osdep.c:48:
include/sysemu/os-win32.h:79:12: error: redundant redeclaration of 'localtime_r' [-Werror=redundant-decls]
 struct tm *localtime_r(const time_t *timep, struct tm *result);
            ^
In file included from include/qemu-common.h:35:0,
                 from util/osdep.c:48:

/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:269:107: note: previous definition of 'localtime_r' was here

This change adds a configure test to see if localtime_r exits, and only enables the QEMU impl if missing. We also re-arrange qemu-common.h try attempt to guarantee that all source files get unistd.h before time.h and thus see the localtime_r/gmtime_r defs.

---
 configure                 | 34 ++++++++++++++++++++++++++++++++++
 include/qemu-common.h     |  4 +++-
 include/sysemu/os-win32.h |  2 ++
 util/oslib-win32.c        |  2 ++
 4 files changed, 41 insertions(+), 1 deletion(-)
configure +34 -0
index 704b34c..322f97b 100755
@@ -1726,6 +1726,37 @@ else
 fi
 
 ##########################################
+# Mingw64 localtime_r/gmtime_r check
+
+if test "$mingw32" = "yes"; then
+    # Some versions of Mingw32/64 lack localtime_r
+    # and gmtime_r entirely
+    #
+    # Some versions of Mingw64 define a macro for
+    # localtime_r/gmtime_r/etc
+    #
+    # Some versions of Ming64 will define functions
+    # for localtime_r/gmtime_r, but only if you have
+    # _POSIX_THREAD_SAFE_FUNCTIONS defined. For fun
+    # though, unistd.h and pthread.h both define
+    # that for you.
+    #
+    # So this #undef localtime_r and #include <unistd.h>
+    # are not in fact redundant
+cat > $TMPC << EOF
+#include <unistd.h>
+#include <time.h>
+#undef localtime_r
+int main(void) { localtime_r(NULL, NULL); return 0; }
+EOF
+    if compile_prog "" "" ; then
+        localtime_r="yes"
+    else
+	localtime_r="no"
+    fi
+fi
+
+##########################################
 # pkg-config probe
 
 if ! has "$pkg_config_exe"; then
@@ -4993,6 +5024,9 @@ fi
 if test "$zero_malloc" = "yes" ; then
   echo "CONFIG_ZERO_MALLOC=y" >> $config_host_mak
 fi
+if test "$localtime_r" = "yes" ; then
+  echo "CONFIG_LOCALTIME_R=y" >> $config_host_mak
+fi
 if test "$qom_cast_debug" = "yes" ; then
   echo "CONFIG_QOM_CAST_DEBUG=y" >> $config_host_mak
 fi
include/qemu-common.h +3 -1
index fb3da6c..e9e3f59 100644
@@ -32,10 +32,12 @@
 #include <strings.h>
 #include <inttypes.h>
 #include <limits.h>
+/* Put unistd.h before time.h as that triggers localtime_r/gmtime_r
+ * function availability on recentish Mingw64 platforms */
+#include <unistd.h>
 #include <time.h>
 #include <ctype.h>
 #include <errno.h>
-#include <unistd.h>
 #include <fcntl.h>
 #include <sys/stat.h>
 #include <sys/time.h>
include/sysemu/os-win32.h +2 -0
index 4035c4f..ba65a25 100644
@@ -73,10 +73,12 @@
 #define siglongjmp(env, val) longjmp(env, val)
 
 /* Missing POSIX functions. Don't use MinGW-w64 macros. */
+#ifndef CONFIG_LOCALTIME_R
 #undef gmtime_r
 struct tm *gmtime_r(const time_t *timep, struct tm *result);
 #undef localtime_r
 struct tm *localtime_r(const time_t *timep, struct tm *result);
+#endif /* CONFIG_LOCALTIME_R */
 
 
 static inline void os_setup_signal_handling(void) {}
util/oslib-win32.c +2 -0
index 730a670..08f5a9c 100644
@@ -95,6 +95,7 @@ void qemu_anon_ram_free(void *ptr, size_t size)
     }
 }
 
+#ifndef CONFIG_LOCALTIME_R
 /* FIXME: add proper locking */
 struct tm *gmtime_r(const time_t *timep, struct tm *result)
 {
@@ -118,6 +119,7 @@ struct tm *localtime_r(const time_t *timep, struct tm *result)
     }
     return p;
 }
+#endif /* CONFIG_LOCALTIME_R */
 
 void qemu_set_block(int fd)
 {
Signature · 1 line
2.4.3
#1 lore raw
Quoted 17 lines · Daniel P. Berrange

CC util/osdep.o
In file included from include/qemu-common.h:48:0,

                 from util/osdep.c:48:
include/sysemu/os-win32.h:77:12: error: redundant redeclaration of 'gmtime_r' [-Werror=redundant-decls]
 struct tm *gmtime_r(const time_t *timep, struct tm *result);
            ^
In file included from include/qemu-common.h:35:0,
                 from util/osdep.c:48:
/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:272:107: note: previous definition of 'gmtime_r' was here
In file included from include/qemu-common.h:48:0,
                 from util/osdep.c:48:
include/sysemu/os-win32.h:79:12: error: redundant redeclaration of 'localtime_r' [-Werror=redundant-decls]
 struct tm *localtime_r(const time_t *timep, struct tm *result);
            ^
In file included from include/qemu-common.h:35:0,
                 from util/osdep.c:48:
Quoted 7 lines

/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:269:107: note: previous definition of 'localtime_r' was here

This change adds a configure test to see if localtime_r exits, and only enables the QEMU impl if missing. We also re-arrange qemu-common.h try attempt to guarantee that all source files get unistd.h before time.h and thus see the localtime_r/gmtime_r defs.

These are only with --enable-werror, right? The patch shouldn't be necessary for 2.4.

Paolo

#2 lore raw
Quoted 22 lines · Paolo Bonzini
Quoted 19 lines · Daniel P. Berrange

CC util/osdep.o
In file included from include/qemu-common.h:48:0,

                 from util/osdep.c:48:
include/sysemu/os-win32.h:77:12: error: redundant redeclaration of 'gmtime_r' [-Werror=redundant-decls]
 struct tm *gmtime_r(const time_t *timep, struct tm *result);
            ^
In file included from include/qemu-common.h:35:0,
                 from util/osdep.c:48:
/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:272:107: note: previous definition of 'gmtime_r' was here
In file included from include/qemu-common.h:48:0,
                 from util/osdep.c:48:
include/sysemu/os-win32.h:79:12: error: redundant redeclaration of 'localtime_r' [-Werror=redundant-decls]
 struct tm *localtime_r(const time_t *timep, struct tm *result);
            ^
In file included from include/qemu-common.h:35:0,
                 from util/osdep.c:48:

/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:269:107: note: previous definition of 'localtime_r' was here

Quoted 8 lines
Quoted 5 lines

This change adds a configure test to see if localtime_r exits, and only enables the QEMU impl if missing. We also re-arrange qemu-common.h try attempt to guarantee that all source files get unistd.h before time.h and thus see the localtime_r/gmtime_r defs.

These are only with --enable-werror, right? The patch shouldn't be necessary for 2.4.

True, these are warnings only, so only fatal if you turn on Werror

Regards,
Daniel

Signature · 4 lines
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|
#3 lore raw
In reply to Paolo Bonzini

Am 31.07.2015 um 14:49 schrieb Paolo Bonzini:

Quoted 24 lines · Daniel P. Berrange
Quoted 16 lines

CC util/osdep.o
In file included from include/qemu-common.h:48:0,

                  from util/osdep.c:48:
include/sysemu/os-win32.h:77:12: error: redundant redeclaration of 'gmtime_r' [-Werror=redundant-decls]
  struct tm *gmtime_r(const time_t *timep, struct tm *result);
             ^
In file included from include/qemu-common.h:35:0,
                  from util/osdep.c:48:
/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:272:107: note: previous definition of 'gmtime_r' was here
In file included from include/qemu-common.h:48:0,
                  from util/osdep.c:48:
include/sysemu/os-win32.h:79:12: error: redundant redeclaration of 'localtime_r' [-Werror=redundant-decls]
  struct tm *localtime_r(const time_t *timep, struct tm *result);
             ^
In file included from include/qemu-common.h:35:0,
                  from util/osdep.c:48:
Quoted 7 lines

/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:269:107: note: previous definition of 'localtime_r' was here

This change adds a configure test to see if localtime_r exits, and only enables the QEMU impl if missing. We also re-arrange qemu-common.h try attempt to guarantee that all source files get unistd.h before time.h and thus see the localtime_r/gmtime_r defs.

Quoted 4 lines

These are only with --enable-werror, right? The patch shouldn't be necessary for 2.4.

Paolo

It isn't necessary for 2.4. There are more severe problems which also seem to be triggered by recent changes in the MinGW-w64 toolchain:

64 bit executables crash at longjmp from generated code, and networking fails because of asynchronous connect calls which were never implemented correctly in the slirp code. I'll try to summarize all that I know in a separate mail later.

I only recently got reports about these problems and think that it is too late to fix them in 2.4.0.

Stefan

#4 lore raw
Quoted 33 lines · Stefan Weil

Am 31.07.2015 um 14:49 schrieb Paolo Bonzini:

Quoted 24 lines · Daniel P. Berrange
Quoted 16 lines

CC util/osdep.o
In file included from include/qemu-common.h:48:0,

                 from util/osdep.c:48:
include/sysemu/os-win32.h:77:12: error: redundant redeclaration of 'gmtime_r' [-Werror=redundant-decls]
 struct tm *gmtime_r(const time_t *timep, struct tm *result);
            ^
In file included from include/qemu-common.h:35:0,
                 from util/osdep.c:48:
/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:272:107: note: previous definition of 'gmtime_r' was here
In file included from include/qemu-common.h:48:0,
                 from util/osdep.c:48:
include/sysemu/os-win32.h:79:12: error: redundant redeclaration of 'localtime_r' [-Werror=redundant-decls]
 struct tm *localtime_r(const time_t *timep, struct tm *result);
            ^
In file included from include/qemu-common.h:35:0,
                 from util/osdep.c:48:
Quoted 7 lines

/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:269:107: note: previous definition of 'localtime_r' was here

This change adds a configure test to see if localtime_r exits, and only enables the QEMU impl if missing. We also re-arrange qemu-common.h try attempt to guarantee that all source files get unistd.h before time.h and thus see the localtime_r/gmtime_r defs.

Quoted 4 lines

These are only with --enable-werror, right? The patch shouldn't be necessary for 2.4.

Paolo

It isn't necessary for 2.4. There are more severe problems which also seem to be triggered by recent changes in the MinGW-w64 toolchain:

Quoted 7 lines

64 bit executables crash at longjmp from generated code, and networking fails because of asynchronous connect calls which were never implemented correctly in the slirp code. I'll try to summarize all that I know in a separate mail later.

I only recently got reports about these problems and think that it is too late to fix them in 2.4.0.

Could you post the info about these problems, or point me to the bug reports. I'm currently trying to get a patchset of mine working on Win32 so need to understand what's broken with Mingw/Win32 currently.

Regards,
Daniel

Signature · 4 lines
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|
#5 lore raw

[Qemu-devel] QEMU 2.4 for Windows - current status

Am 05.08.2015 um 11:52 schrieb Daniel P. Berrange:

Quoted 38 lines · Stefan Weil
Quoted 37 lines

Am 31.07.2015 um 14:49 schrieb Paolo Bonzini:

Quoted 24 lines · Daniel P. Berrange
Quoted 16 lines

CC util/osdep.o
In file included from include/qemu-common.h:48:0,

                  from util/osdep.c:48:
include/sysemu/os-win32.h:77:12: error: redundant redeclaration of 'gmtime_r' [-Werror=redundant-decls]
  struct tm *gmtime_r(const time_t *timep, struct tm *result);
             ^
In file included from include/qemu-common.h:35:0,
                  from util/osdep.c:48:
/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:272:107: note: previous definition of 'gmtime_r' was here
In file included from include/qemu-common.h:48:0,
                  from util/osdep.c:48:
include/sysemu/os-win32.h:79:12: error: redundant redeclaration of 'localtime_r' [-Werror=redundant-decls]
  struct tm *localtime_r(const time_t *timep, struct tm *result);
             ^
In file included from include/qemu-common.h:35:0,
                  from util/osdep.c:48:
Quoted 7 lines

/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:269:107: note: previous definition of 'localtime_r' was here

This change adds a configure test to see if localtime_r exits, and only enables the QEMU impl if missing. We also re-arrange qemu-common.h try attempt to guarantee that all source files get unistd.h before time.h and thus see the localtime_r/gmtime_r defs.

Quoted 4 lines

These are only with --enable-werror, right? The patch shouldn't be necessary for 2.4.

Paolo

It isn't necessary for 2.4. There are more severe problems which also seem to be triggered by recent changes in the MinGW-w64 toolchain:

64 bit executables crash at longjmp from generated code, and networking fails because of asynchronous connect calls which were never implemented correctly in the slirp code. I'll try to summarize all that I know in a separate mail later.

Quoted 8 lines
Quoted 2 lines

I only recently got reports about these problems and think that it is too late to fix them in 2.4.0.

Could you post the info about these problems, or point me to the bug reports. I'm currently trying to get a patchset of mine working on Win32 so need to understand what's broken with Mingw/Win32 currently.

Regards,
Daniel

Sorry for the delay - I have currently only sporadic internet access.

I recently upgraded my build environment from Debian Wheezy to Debian Jessie. Obviously the newer MinGW-w64 packages introduced several problems for QEMU:

* Compiler warnings caused by newer compiler and changes in system include files

* Modified order of include files and definitions. This results in a wrong sigsetjmp call in cpu-exec.c for 64 bit code. That code will crash as soon as there is a longjmp returning from generated TCG code because that code does not support stack unwinding.

* Asynchronous connect call. Obviously TCP connect was synchronous with earlier versions. Now it can also fail, but Windows does not set errno (slirp expects errno code). This results in buggy networking with all Windows versions of QEMU (any connect will fail).

Here are two hacks to fix these problems:

Fix sigsetjmp for w64
http://repo.or.cz/w/qemu/ar7.git/commit/8fa9c07c9a33174905e67589bea6be3e278712cb

slirp: Fix non blocking connect for w32
http://repo.or.cz/w/qemu/ar7.git/commit/b3f21d56ad3f36562d396685de8ff4981af6b805

At least the first patch (maybe also the 2nd one) will also be needed for old versions of QEMU as they don't depend on the QEMU version but on the version of MinGW-w64.

Regards
Stefan

#6 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Quoted 3 lines · Stefan Weil

Fix sigsetjmp for w64
http://repo.or.cz/w/qemu/ar7.git/commit/8fa9c07c9a33174905e67589bea6be3e278712cb

Does this work too?

cpu-exec.c +1 -0
index 75694f3..d882b89 100644
@@ -17,6 +17,7 @@
  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
  */
 #include "config.h"
+#include "qemu-common.h"
 #include "cpu.h"
 #include "trace.h"
 #include "disas/disas.h"
Quoted 2 lines

slirp: Fix non blocking connect for w32
http://repo.or.cz/w/qemu/ar7.git/commit/b3f21d56ad3f36562d396685de8ff4981af6b805

The second looks good, but you do not need the #ifdef at all in theory. If you do, other breakage is looming.

Paolo

#7 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Am 05.08.2015 um 14:49 schrieb Paolo Bonzini:

Quoted 10 lines · Stefan Weil
Quoted 2 lines

Fix sigsetjmp for w64
http://repo.or.cz/w/qemu/ar7.git/commit/8fa9c07c9a33174905e67589bea6be3e278712cb

Does this work too?

cpu-exec.c +0 -0
index 75694f3..d882b89 100644

cpu-exec.c line 23

Quoted 7 lines
cpu-exec.c +1 -0
@@ -17,6 +17,7 @@
   * License along with this library; if not, see <http://www.gnu.org/licenses/>.
   */
  #include "config.h"
+#include "qemu-common.h"
  #include "cpu.h"
  #include "trace.h"
  #include "disas/disas.h"

No, that still gets the wrong definition for setjmp from /usr/share/mingw-w64/include/setjmp.h.

For the future, I'd prefer a solution which restricts the special handling of sigsetjmp to those locations where it is needed. As far as I see, only cpu-exec.c uses sigsetjmp for returns from generated code.

Maybe we can also add support for stack unwinding to the generated code, so no special hack for w64 will be needed anymore.

Stefan

#8 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

In reply to Paolo Bonzini

Am 05.08.2015 um 14:49 schrieb Paolo Bonzini:

Quoted 1 line · Stefan Weil

[...]

Quoted 7 lines
Quoted 2 lines

slirp: Fix non blocking connect for w32
http://repo.or.cz/w/qemu/ar7.git/commit/b3f21d56ad3f36562d396685de8ff4981af6b805

The second looks good, but you do not need the #ifdef at all in theory. If you do, other breakage is looming.

Paolo

The new code was this:

          if ((tcp_fconnect(so) == -1) &&
#if defined(_WIN32)
              socket_error() != WSAEWOULDBLOCK
#else
              (errno != EINPROGRESS) && (errno != EWOULDBLOCK)
#endif
          ) {

Without the preprocessor conditionals, Linux users would get an undefined WSAEWOULDBLOCK.

And QEMU defines a macro EWOULDBLOCK which differs from the MinGW-w64 definition, at least in my Debian Jessie cross environment (10035 != 140):

include/sysemu/os-win32.h:

 # define EWOULDBLOCK  WSAEWOULDBLOCK

/usr/share/mingw-w64/include/pthread.h:
#define EWOULDBLOCK 140
/usr/share/mingw-w64/include/psdk_inc/_wsa_errnos.h:
#define WSAEWOULDBLOCK (WSABASEERR + 35 )

That's only a problem if we get pthread.h when compiling for Windows. Can we be sure that this won't happen? It looks like Windows + MinGW(-w64) is a moving target which is not so easy to handle.

Maybe it would be better to change tcp_fconnect() to return an ANSI error code (with a conversion from Windows error codes in that function).

Or we introduce a wrapper for connect(), qemu_connect().

Stefan

#9 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

In reply to Stefan Weil
Quoted 5 lines · Stefan Weil

Fix sigsetjmp for w64
http://repo.or.cz/w/qemu/ar7.git/commit/8fa9c07c9a33174905e67589bea6be3e278712cb

I tried to apply your patch to my branch and I got:

Running QEMU make install...
CC gnuarmeclipse-softmmu/cpu-exec.o
/Host/Work/qemu/gnuarmeclipse-qemu.git/cpu-exec.c:37:0: warning: "sigsetjmp" redefined
#define sigsetjmp(env, savesigs) _setjmp(env, NULL)
^
In file included from /Host/Work/qemu/gnuarmeclipse-qemu.git/include/qemu-common.h:47:0,
from /Host/Work/qemu/gnuarmeclipse-qemu.git/target-arm/cpu.h:39,
from /Host/Work/qemu/gnuarmeclipse-qemu.git/cpu-exec.c:20:
/Host/Work/qemu/gnuarmeclipse-qemu.git/include/sysemu/os-win32.h:72:0: note: this is the location of the previous definition
#define sigsetjmp(env, savemask) setjmp(env)
^

I did not check the latest version in the repository (waiting for 2.4), but could you verify, maybe a better location for your patch would be in os-win32.h?

Regards,

Liviu

#10 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Quoted 22 lines · Liviu Ionescu
Quoted 5 lines · Stefan Weil

Fix sigsetjmp for w64
http://repo.or.cz/w/qemu/ar7.git/commit/8fa9c07c9a33174905e67589bea6be3e278712cb

I tried to apply your patch to my branch and I got:

Running QEMU make install...
CC gnuarmeclipse-softmmu/cpu-exec.o
/Host/Work/qemu/gnuarmeclipse-qemu.git/cpu-exec.c:37:0: warning: "sigsetjmp" redefined
#define sigsetjmp(env, savesigs) _setjmp(env, NULL)
^
In file included from /Host/Work/qemu/gnuarmeclipse-qemu.git/include/qemu-common.h:47:0,
from /Host/Work/qemu/gnuarmeclipse-qemu.git/target-arm/cpu.h:39,
from /Host/Work/qemu/gnuarmeclipse-qemu.git/cpu-exec.c:20:
/Host/Work/qemu/gnuarmeclipse-qemu.git/include/sysemu/os-win32.h:72:0: note: this is the location of the previous definition
#define sigsetjmp(env, savemask) setjmp(env)
^

Quoted 3 lines

I did not check the latest version in the repository (waiting for 2.4), but could you verify, maybe a better location for your patch would be in os-win32.h?

os-win32.h has the following already:

#if defined(_WIN64)

# undef setjmp
# define setjmp(env) _setjmp(env, NULL)
#endif
#define sigjmp_buf jmp_buf
#define sigsetjmp(env, savemask) setjmp(env)

And Stefan's patch has this in cpu-exec.c:

#if defined(_WIN64)
#define sigsetjmp(env, savesigs) _setjmp(env, NULL) #endif

The above should not be necessary if the definition of sigsetjmp is already picked up from os-win32.h.

Since os-win32.h is usually included through qemu-common.h, my guess was that cpu-exec.c was missing an inclusion of qemu-common.h. However, all target-*/cpu.h files include qemu-common.h, so I am not sure why things break for Stefan...

Paolo

#11 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Quoted 4 lines · Paolo Bonzini

... I am not sure why things break for Stefan...

I confirm Stefan's conclusion, neither in my configuration adding

#include "qemu-common.h"

... in cpu-exec.c makes any difference.

however adding:

#if defined(_WIN64)
#ifdef sigsetjmp
#undef sigsetjmp
#endif
#define sigsetjmp(env, savesigs) _setjmp(env, NULL) #endif

... fixes the problem, my custom QEMU happily blinks the LEDs on Win 8.1 64-bits (see below).

perhaps a headers check would be helpful, such mysterious behaviours usually back fire at a certain point.

regards,

Liviu

GNU ARM Eclipse 64-bits QEMU v2.3.50 (qemu-system-gnuarmeclipse.exe).
Board: 'STM32F4-Discovery' (ST Discovery kit for STM32F407/417 lines).
Device: 'STM32F407VG' (Cortex-M4 r0p0, MPU), Flash: 1024 kB, RAM: 128 kB. Command line: 'f4' (2 bytes).
Cortex-M4 r0p0 core initialised.
GDB Server listening on: 'tcp::1234'...
Cortex-M4 r0p0 core reset.
... connection accepted from 127.0.0.1.

PRIGROUP unimplemented
Execute 'mon system_reset'.
Cortex-M4 r0p0 core reset.

main(argc=1, argv=["f4"]);
Hello ARM World!
System clock: 168000000Hz
Standard output message.
Standard error message.
[Green LED On]
[Green LED Off]
Second 1
[Green LED On]
[Green LED Off]
Second 2
[Green LED On]
[Green LED Off]
Second 3
[Green LED On]
[Green LED Off]
Second 4
[Green LED On]
[Green LED Off]
Second 5
QEMU exit(0)

#12 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Am 05.08.2015 um 20:39 schrieb Liviu Ionescu:

Quoted 18 lines
Quoted 3 lines · Paolo Bonzini

... I am not sure why things break for Stefan...

I confirm Stefan's conclusion, neither in my configuration adding

#include "qemu-common.h"

... in cpu-exec.c makes any difference.

however adding:

#if defined(_WIN64)
#ifdef sigsetjmp
#undef sigsetjmp
#endif
#define sigsetjmp(env, savesigs) _setjmp(env, NULL) #endif

Quoted 8 lines

... fixes the problem, my custom QEMU happily blinks the LEDs on Win 8.1 64-bits (see below).

perhaps a headers check would be helpful, such mysterious behaviours usually back fire at a certain point.

regards,

Liviu

http://qemu.weilnetz.de/test/cpu-exec.i shows the result of the C preprocessor:

cpu-exec.c gets QEMU's os-win32.h with our definition of setjmp early, but the system header file setjmp.h is included later, and that file re-defines our definitions. Including setjmp.h from os-win32.h would solve the problem, but I think there is a better solution.

I am planning to remove the special definitions for _WIN64 from os-win32.h and add them to cpu-exec.c, similar to the code above (which can be shortened a little)
but with some comment lines added.

As I already said, this modification is needed for all versions of QEMU, and it will stay unfixed in 2.4.0 which is nearly finished.

Regards
Stefan

#13 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Quoted 5 lines · Stefan Weil

... As I already said, this modification is needed for all versions of QEMU, and it will stay unfixed in 2.4.0 which is nearly finished.

your patch is already in my branch, and I already publicly released GNU ARM Eclipse QEMU for Windows 64-bits, so no problem, take your time and work on a solution you consider appropriate.

regards,

Liviu

#14 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

In reply to Stefan Weil

Am 05.08.2015 um 22:30 hat Stefan Weil geschrieben:

Quoted 37 lines

Am 05.08.2015 um 20:39 schrieb Liviu Ionescu:

Quoted 18 lines
Quoted 3 lines · Paolo Bonzini

... I am not sure why things break for Stefan...

I confirm Stefan's conclusion, neither in my configuration adding

#include "qemu-common.h"

... in cpu-exec.c makes any difference.

however adding:

#if defined(_WIN64)
#ifdef sigsetjmp
#undef sigsetjmp
#endif
#define sigsetjmp(env, savesigs) _setjmp(env, NULL) #endif

Quoted 8 lines

... fixes the problem, my custom QEMU happily blinks the LEDs on Win 8.1 64-bits (see below).

perhaps a headers check would be helpful, such mysterious behaviours usually back fire at a certain point.

regards,

Liviu

http://qemu.weilnetz.de/test/cpu-exec.i shows the result of the C preprocessor:

cpu-exec.c gets QEMU's os-win32.h with our definition of setjmp early, but the system header file setjmp.h is included later, and that file re-defines our definitions. Including setjmp.h from os-win32.h would solve the problem, but I think there is a better solution.

Quoted 4 lines

I am planning to remove the special definitions for _WIN64 from os-win32.h and add them to cpu-exec.c, similar to the code above (which can be shortened a little)
but with some comment lines added.

What happens with the callers outside of cpu-exec.c then? Fixing the problem inside os-win32.h seems more robust to me.

Kevin

#15 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Am 06.08.2015 um 10:44 schrieb Kevin Wolf:

Quoted 37 lines

Am 05.08.2015 um 22:30 hat Stefan Weil geschrieben:

Quoted 36 lines

Am 05.08.2015 um 20:39 schrieb Liviu Ionescu:

Quoted 18 lines
Quoted 3 lines · Paolo Bonzini

... I am not sure why things break for Stefan...

I confirm Stefan's conclusion, neither in my configuration adding

#include "qemu-common.h"

... in cpu-exec.c makes any difference.

however adding:

#if defined(_WIN64)
#ifdef sigsetjmp
#undef sigsetjmp
#endif
#define sigsetjmp(env, savesigs) _setjmp(env, NULL) #endif

Quoted 8 lines

... fixes the problem, my custom QEMU happily blinks the LEDs on Win 8.1 64-bits (see below).

perhaps a headers check would be helpful, such mysterious behaviours usually back fire at a certain point.

regards,

Liviu

http://qemu.weilnetz.de/test/cpu-exec.i shows the result of the C preprocessor:

cpu-exec.c gets QEMU's os-win32.h with our definition of setjmp early, but the system header file setjmp.h is included later, and that file re-defines our definitions. Including setjmp.h from os-win32.h would solve the problem, but I think there is a better solution.

Quoted 8 lines
Quoted 4 lines

I am planning to remove the special definitions for _WIN64 from os-win32.h and add them to cpu-exec.c, similar to the code above (which can be shortened a little)
but with some comment lines added.

What happens with the callers outside of cpu-exec.c then? Fixing the problem inside os-win32.h seems more robust to me.

Kevin

(sig)setjmp/(sig)longjmp without TCG generated code on the stack does not need special handling because stack unwinding works for compiled normal C code.

We only have a problem on 64 bit Windows with code generated at run time by TCG because stack unwinding does not work there, so a siglongjmp which tries to exit from that code will crash QEMU when trying to unwind the call stack.

As far as I see, the only siglongjmp exits from TCG generated code are in cpu-exec.c (functions cpu_loop_exit and cpu_resume_from_signal). The matching sigsetjmp is also in cpu-exec.c, and here we have to disable stack unwinding by calling _setjmp(cpu->jmp_env, NULL).

Stefan

#16 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Quoted 8 lines · Stefan Weil

(sig)setjmp/(sig)longjmp without TCG generated code on the stack does not need special handling because stack unwinding works for compiled normal C code.

We only have a problem on 64 bit Windows with code generated at run time by TCG because stack unwinding does not work there, so a siglongjmp which tries to exit from that code will crash QEMU when trying to unwind the call stack.

I suppose we could fix that, for 2.5, anyway.
It would appear that RtlAddFunctionTable is
the proper interface.

It would probably also help debugging just as
much as adding the elf unwind info did.

Quoted 5 lines

As far as I see, the only siglongjmp exits from TCG generated code are in cpu-exec.c (functions cpu_loop_exit and cpu_resume_from_signal). The matching sigsetjmp is also in cpu-exec.c, and here we have to disable stack unwinding by calling _setjmp(cpu->jmp_env, NULL).

That sounds right.

r~

#17 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

In reply to Stefan Weil
Quoted 13 lines · Stefan Weil

I recently upgraded my build environment from Debian Wheezy to Debian Jessie. Obviously the newer MinGW-w64 packages introduced several problems for QEMU:

* Compiler warnings caused by newer compiler and changes in system include files

* Modified order of include files and definitions. This results in a wrong sigsetjmp call in cpu-exec.c for 64 bit code. That code will crash as soon as there is a longjmp returning from generated TCG code because that code does not support stack unwinding.

Quoted 5 lines

* Asynchronous connect call. Obviously TCP connect was synchronous with earlier versions. Now it can also fail, but Windows does not set errno (slirp expects errno code). This results in buggy networking with all Windows versions of QEMU (any connect will fail).

Could somebody add some notes to the 'Known issues' section of the changelog summarising the problems, please (whether it affects all Windows build environments or only some, whether this is a regression or not, etc):
http://wiki.qemu.org/ChangeLog/2.4#Known_issues

thanks
-- PMM

#18 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Am 10.08.2015 um 12:25 schrieb Peter Maydell:

Quoted 9 lines

Could somebody add some notes to the 'Known issues' section of the changelog summarising the problems, please (whether it affects all Windows build environments or only some, whether this is a regression or not, etc):
http://wiki.qemu.org/ChangeLog/2.4#Known_issues

thanks
-- PMM

Done. I also added the BIOS regression and the missing GTK ui translations.

Stefan

#19 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Quoted 10 lines · Stefan Weil
Quoted 8 lines

Could somebody add some notes to the 'Known issues' section of the changelog summarising the problems, please (whether it affects all Windows build environments or only some, whether this is a regression or not, etc):
http://wiki.qemu.org/ChangeLog/2.4#Known_issues

thanks
-- PMM

Done. I also added the BIOS regression

Isn't it harmless anyway?

Paolo

#20 lore raw

Re: [Qemu-devel] QEMU 2.4 for Windows - current status

Am 10.08.2015 um 16:00 schrieb Paolo Bonzini:

Quoted 9 lines · Stefan Weil
Quoted 6 lines
Quoted 6 lines

Could somebody add some notes to the 'Known issues' section of the changelog summarising the problems, please (whether it affects all Windows build environments or only some, whether this is a regression or not, etc):
http://wiki.qemu.org/ChangeLog/2.4#Known_issues

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

thanks
-- PMM

Done. I also added the BIOS regression

Isn't it harmless anyway?

Paolo

I assume it is harmless for most users who don't use much BIOS functionality.

Nevertheless there are several outl function calls with wrong argument order. Those calls won't work as expected. I don't know how important this is.

My QEMU tree includes trace messages for unaligned i/o, so users of my precompiled QEMU for Windows will see a message during system boot.

Regards
Stefan