Hello,

> It's really great to see all of this activity!
> Thanks Tristan!

Well, thank you for reloading ghdl!

> A couple of questions if you don't mind; things I don't understand
> and
> so on.
> 
> ------- from 77dde4 ---------------------------
> sem_assocs: add missing set_base_name on formal.
> Did you just notice this was missing (I'd never have seen it!) or is
> it part of another commit?

In fact, I'd like to rework change set 156 & 157: crashes in
Get_Kind means there is an issue before.
(I will revert the Get_Kind change, as it isn't anymore necessary
 for the reported issues.)

> ------- from 77dde4 ---------------------------
> https://sourceforge.net/p/ghdl-updates/code/ci/77dde45d5c939de69f57658cc7b23f4609223802/
> 
>        --  Extract high or low bound of RANGE_VAR, which must be
>        stable.
>        --  Put the result into RES.
> -      procedure Range_To_High_Low
> -        (Range_Var : Mnode; Res : O_Dnode; Is_High : Boolean)
> +      function Range_To_High_Low
> +        (Range_Var : Mnode; Range_Type : Iir; Is_High : Boolean)
> +        return Mnode
> 
> The comment looks out of date?

Yes.

> ------- from 1f3238 and c19ce6 ---------------------------
> 
> Adding an "ENTITY" option to various tests ...
> 
> I was in the process of trying to get the entity automatically, e.g.
> from recent tests (and checking it's named in an architecture
> specification) to limit fiddling with the .exp files.
> 
> Is that still worthwhile or shall I abandon the attempt and follow
> this approach?

It is up to you. I have modified get_entities (should now be renamed ?)
to print the last entity of the file, which is the one at the top of
the design. And I added ENTITY when the design must be elaborated from
a configuration.

I am not sure it is worth trying to improve that.  Not very important
nor urgent.

> (I have added a STOP option to terminate simulation after time T, to
> reduce the number of tests we have to fix for infinite sim length)

Good, might be useful.

>  -------------------------------------------------------------
> 
> Also I'll add a note to the downstream Debian bug reports that have
> now been fixed.

Great.  I did the same for most gna.org reports.

I also plan to revert changeset 135: in fact there was a missing range
check during sem. The range is an integer range, and the right bound
is clearly out of range.

I also have a question about one of your change:

--- a/sem_names.adb
+++ b/sem_names.adb
@@ -1980,8 +1980,8 @@ package body Sem_Names is
                end;
                if Res = Null_Iir then
                   Error_Msg_Sem
-                    ("prefix is neither a function name "
-                     & "nor can it be sliced or indexed", Name);
+                    ("No overloaded subprogram found matching "
+                       & Disp_Node (Prefix_Name), Name);
                end if;
             when Iir_Kinds_Function_Declaration =>
                Add_Result (Res, Sem_As_Function_Call (Prefix_Name,

I wonder wether the message is an improvement or not, as the prefix may
not be a function. But I suppose you'd like to display the prefix to
make the message clearer ?

Regards,
Tristan.


_______________________________________________
Ghdl-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/ghdl-discuss

Reply via email to