gEDA: Re: struct t_vpi_systf_data

[Stephen Williams <steve@icarus.com>]

Larry Doolittle wrote:
> Steve -
>
> Icarus' vpi_user.h contains the following declaration:
>
> typedef struct t_vpi_systf_data {
>       PLI_INT32 type;
>       PLI_INT32 subtype;
>       char      *tfname;
>       PLI_INT32 (*calltf)(char*);
>       PLI_INT32 (*compiletf)(char*);
>       PLI_INT32 (*sizetf)();
>       char      *user_data;
> } s_vpi_systf_data, *p_vpi_systf_data;
>
> and a typical code snippet I use to fill in this structure is
>
>       s_vpi_systf_data tf_data;
>       tf_data.type      = vpiSysTask;
>       tf_data.tfname    = "$llrf_sysmodel";
>       tf_data.calltf    = sysmodel_calltf;
>       tf_data.compiletf = sysmodel_compiletf;
>       tf_data.sizetf    = 0;
>       tf_data.user_data = "$llrf_sysmodel";
>
> I recently turned gcc's warning level up another few notches.
> When I add -Wwrite-strings to the mix, gcc barks at the above:
>
> warning: assignment discards qualifiers from pointer target type
>
> This actually makes sense: how am I supposed to know that the routine
> that manipulates this data structure (vpi_register_systf) never modifies
> foo.tfname or foo.user_data?  If it _might_, I need to strdup().
> If it is guaranteed not to, the structure member declarations for
> tfname and user_data need "const" added.
>
> O Verilog Master, which is the one true path?
The IEEE1364 folks seem to have lacked the practical C programming
experience needed to get things like this right. In fact, the standard
declares tfname to be PLI_BYTE*, which is sheer nonsense. They clearly
did a global substitute on the header file. We will do the right thing
and call it "const char*" which is backwards compatible with 1364.

I've edited my copy of vpi_user.h, so the change will make its
way to CVS in due course. I will consider filing this as a change
request with the IEEE, but it might not fly due to some ancient
C compilers not supporting the const keyword.

The user_data member is a little bit more tricky. It too is not
modified by the VPI. Heck, it really should be "const void*" but
given the way users would tend to use it, that change may cause
more warnings then it fixes.
--
Steve Williams                "The woods are lovely, dark and deep.
steve at icarus.com           But I have promises to keep,
http://www.icarus.com         and lines to code before I sleep,
http://www.picturel.com       And lines to code before I sleep."