sem_ch3.adb (Analyze_Object_Declaration): Avoid type Any_Access in unresolved…

sem_ch3.adb (Analyze_Object_Declaration): Avoid type Any_Access in unresolved initial value of "null"...

2008-08-05  Bob Duff  <duff@adacore.com>

	* sem_ch3.adb (Analyze_Object_Declaration): Avoid type Any_Access in
	unresolved initial value of "null", because it causes implicitly
	generated "=" operators to be ambiguous, and because this type should
	not be passed to gigi.

From-SVN: r138693

gcc/ada/ChangeLog

+2008-08-05  Jose Ruiz  <ruiz@adacore.com>
+
+	* init.c (__gnat_install_handler for linux): If we are building the
+	Xenomai run time then we need to do two additional things: avoid
+	memory swapping and transform the Linux environment task into a native
+	Xenomai task.
+
+	* gcc-interface/Makefile.in (LIBGNAT_TARGET_PAIRS for xenomai run
+	time): Use interface to Xenomai native skin and avoid linux-specific
+	way of setting CPU affinity.
+	(EH_MECHANISM for the xenomai run time): Use sjlj exception mechanism.
+
+2008-08-05  Bob Duff  <duff@adacore.com>
+
+	* checks.ads: Minor comment fix
+
+2008-08-05  Thomas Quinot  <quinot@adacore.com>
+
+	* g-sercom.adb, g-sercom.ads, g-sercom-mingw.adb,
+	g-sercom-linux.adb (Data_Bits): Change literals B7 and B8 to CS7 and
+	CS8.
+
+2008-08-05  Robert Dewar  <dewar@adacore.com>
+
+	* mlib.adb: Minor code reorganization
+	Minor reformatting
+
+	* make.adb: Minor reformatting
+
+	* prj-attr.ads: Minor reformatting
+
+	* s-os_lib.adb: Minor reformatting
+
+	* s-fileio.adb: Minor code reorganization
+	Minor reformatting
+
+	* prj.ads: Minor reformatting
+
+2008-08-05  Bob Duff  <duff@adacore.com>
+
+	* sem_ch3.adb (Analyze_Object_Declaration): Avoid type Any_Access in
+	unresolved initial value of "null", because it causes implicitly
+	generated "=" operators to be ambiguous, and because this type should
+	not be passed to gigi.
+
 2008-08-05  Vincent Celier  <celier@adacore.com>
 
 	* mlib.adb: Update comments.

gcc/ada/sem_ch3.adb

@@ -2550,6 +2550,15 @@ package body Sem_Ch3 is
          Set_Etype (Id, T);
          Resolve (E, T);
 
+         --  If E is "null", and has been replaced by an
+         --  N_Raise_Constraint_Error node (which was marked already-analyzed),
+         --  we need to set the type to something other than Any_Access in
+         --  order to keep gigi happy.
+
+         if Etype (E) = Any_Access then
+            Set_Etype (E, T);
+         end if;
+
          --  If the object is an access to variable, the initialization
          --  expression cannot be an access to constant.