On 30/09/15 23:48, Joris van Rantwijk wrote:
On 2015-09-30, Tristan Gingold wrote:
On 30/09/15 13:04, Joris van Rantwijk wrote:
When building the LLVM backend on i386, I get compiler errors.
Should be fixed now. Sorry, I didn't try this configuration after
the changes in config/
Yes, fixed.
Thanks.
Next problem. So far I see this only with LLVM backend on i386.
During building of the IEEE library, GHDL fails:
../../../../ghdl_llvm -a --GHDL1=ghdl1-llvm -g --std=87 -P../std --work=ieee
../../src/ieee/numeric_bit-body.v87
[...]
On casual inspection of trans-chap2.adb, Translate_Subprogram_Body, it appears
that the variable Is_Ortho_Func may be used uninitialized.
Ah, thanks. Can you try the following patch.
Which distribution and version of linux are you using ?
Tristan.
diff -r 019e94db0d22 src/vhdl/translate/trans-chap2.adb
--- a/src/vhdl/translate/trans-chap2.adb Wed Sep 30 20:59:13 2015 +0200
+++ b/src/vhdl/translate/trans-chap2.adb Thu Oct 01 19:58:42 2015 +0200
@@ -368,10 +368,13 @@
Get_Kind (Spec) = Iir_Kind_Procedure_Declaration
and then Get_Suspend_Flag (Spec);
+ -- True if the subprogram is translated to a function in ortho.
+ Is_Ortho_Func : constant Boolean := Is_Subprogram_Ortho_Function
(Spec);
+
Old_Subprogram : Iir;
Mark : Id_Mark_Type;
Final : Boolean;
- Is_Ortho_Func : Boolean;
+
-- Set for a public method. In this case, the lock must be
acquired
-- and retained.
@@ -596,14 +599,11 @@
-- If finalization is required and if the subprogram is a function,
-- create a variable for the result.
- if Final or Is_Prot then
- Is_Ortho_Func := Is_Subprogram_Ortho_Function (Spec);
- if Is_Ortho_Func then
- New_Var_Decl
- (Info.Subprg_Result, Get_Identifier ("RESULT"),
- O_Storage_Local,
- Get_Ortho_Type (Get_Return_Type (Spec), Mode_Value));
- end if;
+ if (Final or Is_Prot) and Is_Ortho_Func then
+ New_Var_Decl
+ (Info.Subprg_Result, Get_Identifier ("RESULT"),
+ O_Storage_Local,
+ Get_Ortho_Type (Get_Return_Type (Spec), Mode_Value));
end if;
-- If finalization is required, create a dummy loop around the
_______________________________________________
Ghdl-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/ghdl-discuss
