Обсуждение: Code inconsistency in convert.c for guid -> string conversion

Hi all,

While working on the code of odbc, I noticed an inconsistency in
convert.c, found by compiling the code with -Werror.

convert.c:4466: error: format '%08lX' expects type 'long unsigned
int', but argument 4 has type 'DWORD'

By looking closer at the code, the behavior in the code for guid ->
string conversion is inconsistent with char2guid using an sscan for
the similar reverse operation. I am attaching a patch fixing that.

On another note, I found as well the following warning:
dlg_specific.c:1491: error: the address of 'conn_settings' will always
evaluate as 'true'
All of this was reproduced using 09.03.0200.

Regards,
--
Michael

On Mon, Mar 3, 2014 at 11:38 AM, Michael Paquier
<michael.paquier@gmail.com> wrote:
> On another note, I found as well the following warning:
> dlg_specific.c:1491: error: the address of 'conn_settings' will always
> evaluate as 'true'
Sorry for the imprecision, this warning is caused by this code:
        if ('\0' != conn_settings[0])
            STR_TO_NAME(comval->conn_settings, conn_settings);
--
Michael

On 03/03/2014 04:41 AM, Michael Paquier wrote:
> On Mon, Mar 3, 2014 at 11:38 AM, Michael Paquier
> <michael.paquier@gmail.com> wrote:
>> On another note, I found as well the following warning:
>> dlg_specific.c:1491: error: the address of 'conn_settings' will always
>> evaluate as 'true'
> Sorry for the imprecision, this warning is caused by this code:
>          if ('\0' != conn_settings[0])
>              STR_TO_NAME(comval->conn_settings, conn_settings);

Thanks, committed.

- Heikki

On 03/03/2014 04:38 AM, Michael Paquier wrote:
> On another note, I found as well the following warning:
> dlg_specific.c:1491: error: the address of 'conn_settings' will always
> evaluate as 'true'

Yeah, that has bothered me for a long time, but at a quick glance I
haven't seen a very good way to fix it. Clearly the "conn_settings ==
NULL" test is unnecessary here, but the free() call that STR_TO_NAME
also does is not, so we can't just replace the macro with an strcpy.

Hmm, now that I look at it, we have a macro STRX_TO_NAME which is the
same but without the NULL check. So we can just use that; that appears
to be how it's used elsewhere too.

So, fixed.

- Heikki

On Tue, Mar 4, 2014 at 4:44 AM, Heikki Linnakangas
<hlinnakangas@vmware.com> wrote:
> On 03/03/2014 04:38 AM, Michael Paquier wrote:
>>
>> On another note, I found as well the following warning:
>> dlg_specific.c:1491: error: the address of 'conn_settings' will always
>> evaluate as 'true'
>
>
> Yeah, that has bothered me for a long time, but at a quick glance I haven't
> seen a very good way to fix it. Clearly the "conn_settings == NULL" test is
> unnecessary here, but the free() call that STR_TO_NAME also does is not, so
> we can't just replace the macro with an strcpy.
>
> Hmm, now that I look at it, we have a macro STRX_TO_NAME which is the same
> but without the NULL check. So we can just use that; that appears to be how
> it's used elsewhere too.
Thanks, I was not sure how to fix this one as well.
--
Michael