https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66205
Bug ID: 66205
Summary: gnatbind generates invalid code when finalization is
enabled in restricted runtime
Product: gcc
Version: 4.9.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: ada
Assignee: unassigned at gcc dot gnu.org
Reporter: simon at pushface dot org
Target Milestone: ---
Created attachment 35567
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35567&action=edit
Patch to gcc/ada/bindgen.adb
This is on 4.9.1 compiled for arm-eabi on Mac OS X Mavericks (Darwin 13).
Please note that the same problem occurs with 5.1.0.
package System has
pragma Profile (Ravenscar);
pragma Restrictions (No_Enumeration_Maps);
pragma Restrictions (No_Exception_Propagation);
pragma Restrictions (No_Recursion);
and the result of a gprbuild is
gprbuild -p -P testbed
gprbind testbed.bexch
arm-eabi-gnatbind testbed.ali
arm-eabi-gcc -c b__testbed.adb
b__testbed.adb:31:14: "Is_Elaborated" is undefined (more references follow)
gprbind: compilation of binder generated file failed
gprbuild: unable to bind testbed.adb
which happens because bindgen.adb has at :2497
if not Suppress_Standard_Library_On_Target then
WBI (" Is_Elaborated : Boolean := False;");
end if;
but at :408 in the generation of adafinal
if not CodePeer_Mode then
WBI (" if not Is_Elaborated then");
WBI (" return;");
WBI (" end if;");
WBI (" Is_Elaborated := False;");
end if;
I’ve worked out a patch for this, but I’m confused about when to use
Suppress_Standard_Library_On_Target vs Configurable_Run_Time_On_Target vs
Configurable_Run_Time_Mode to determine that adafinal will never actually be
called & hence needn’t do anything.
Not sure whether __gnat_initialize, __gnat_finalize are expected (they aren’t
in the AdaCore arm-eabi libraries), so I’ve suppressed them too.
