Ghdl: GHDL able to define a string with a range outside type "string" as in (4 downto 0)

[sckoarn]

Description
GHDL is able to process a string definition which is outside the predefined range of string (natural range <>) which does not include 0

Expected behaviour
In the case of the EDA Playground example below, it should fail at run time when a dynamic string with range that includes 0 is attempted.

How to reproduce?
https://www.edaplayground.com/x/5smV

NOTE: :file: and :image: identifiers are specific to issue-runner. We suggest to use these, since it allows continuous integration workflows to automatically test the MWE. Using ghdl/ghdl:* docker images to run the MWEs ensures that the latest available GHDL is used.

NOTE: Large files can be uploaded one-by-one or in a tarball/zipfile.

Context
Please, provide the following information:

• OS:
• Origin:
  
  
  
  • [ ] Package manager: version
  
  
  • [ ] Released binaries: tarball_url
  
  
  • [ ] Built from sources: commit SHA
  
  
  • 
    
    
    • EDA Playground
    
    
    
  
  

If a GHDL Bug occurred block is shown in the log, please paste it here:

******************** GHDL Bug occurred ***************************
Please report this bug on https://github.com/ghdl/ghdl/issues
...
******************************************************************

Additional context
Add any other context about the problem here. If applicable, add screenshots to help explain your problem.

[eine]

FTR:

entity ent is
end;

architecture arch of ent is
  signal strt: string(0 to 4);
  signal strd: string(4 downto 0);
begin
process
  variable strvt: string(0 to 4);
  variable strvd: string(4 downto 0);
begin
report "HELLO" severity note;
wait;
end process;
end;

# ghdl -a test.vhd 
test.vhd:5:23:warning: static range violates bounds [-Wruntime-error]
test.vhd:5:23:warning: static range violates bounds [-Wruntime-error]
test.vhd:6:23:warning: static range violates bounds [-Wruntime-error]
test.vhd:6:23:warning: static range violates bounds [-Wruntime-error]
test.vhd:9:26:warning: static range violates bounds [-Wruntime-error]
test.vhd:9:26:warning: static range violates bounds [-Wruntime-error]
test.vhd:10:26:warning: static range violates bounds [-Wruntime-error]
test.vhd:10:26:warning: static range violates bounds [-Wruntime-error]

---

entity ent is
end;

architecture arch of ent is
begin
process
  function bv2str(bv: bit_vector) return string is
    variable st_out : string(bv'high downto bv'low);
    --variable st_out : string(bv'range); -- produces the same result
  begin
    for i in bv'range loop
      if(bv(i) = '0') then
        st_out(i) := '0';
      else
        st_out(i) := '1';
      end if;
    end loop;
    return st_out;
  end function;
  variable bitv : bit_vector(4 downto 0);
begin
report bv2str(bitv);
wait;
end process;
end;

# ./ent
test.vhd:22:1:@0ms:(report note): 00000

---

entity ent is
end;

architecture arch of ent is
begin
process
  function bv2str(bv: bit_vector) return string is
    variable st_out : string(4 downto 0);
  begin
    for i in bv'range loop
      if(bv(i) = '0') then
        st_out(i) := '0';
      else
        st_out(i) := '1';
      end if;
    end loop;
    return st_out;
  end function;
  variable bitv : bit_vector(4 downto 0);
begin
report "HELLO" severity note;
report bv2str(bitv);
wait;
end process;
end;

# ghdl -a test.vhd
test.vhd:8:30:warning: static range violates bounds [-Wruntime-error]
test.vhd:8:30:warning: static range violates bounds [-Wruntime-error]

[alemuller]

This is the expected behavior.

The string type is defines as:

type STRING is array (POSITIVE range <>) of CHARACTER;

The zero index violates the positive range constraint.

[eine]

@alemuller that is the point. When variable bitv : bit_vector(4 downto 0); is passed to bv2str(bitv), variable st_out : string(bv'high downto bv'low); is variable st_out : string(4 downto 0);, which we think it should produce a range violation error. However, that's not the case.

The first and third example of my previous comment show that in some contexts GHDL produces a warning/error, as expected. However, in the second example (which is @sckoarn's), no warning/error is produced, and we would expect it.

[alemuller]

I see.

MWE:

entity ent is
    generic (g : integer := 4);
end;

architecture arch of ent is
    signal str : string (g downto 0) := (others=>'a');
begin
    process
    begin
        report str;
        wait;
    end process;
end;

I'm note sure what the spec says about using globally static expression to define range boundary.