Mailing List Archive

Do you have glibc on your machine ?
On 14 Jan 2016 20:27, "K. Peter" <kp@dyndn.es> wrote:

> Hi,
>
> I was coming back to an issue I saw in the past but didn't anything. I
> think it is related to all qmail derivates. At least qmail-1.03 and
> netqmail-1.06 are affected as s/qmail-3.01 too.
>
> During make several times gcc creates a warning for .c files which uses
> the SUBSTDIO_FDBUF macro, defined in substdio.h. This example reproduces it:
>
> # gcc -fsyntax-only subfderr.c
> subfderr.c:8:1: warning: initialization from incompatible pointer type
> static substdio it = SUBSTDIO_FDBUF(write,2,subfd_errbuf,256);
> ^
> subfderr.c:8:1: warning: (near initialization for ‘it.op’)
>
>
> With my rudimental C expierience I would expect that a cast have to be
> done, but maybe I'm wrong. So the questions are
>
> 1. Does anybody have had the same issue and did solve it? What is the
> solution?
> 2. Could anybody provide a solution or point me in the right direction?
> 3. In case a pointer cast is needed, where have it be done?
>
> System is an up-to-date Gentoo with gcc-4.9.3.
>
> Any help is appreciated.
>
> thanks and regards
> Kai
>
> --
> -⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠
> Dyn@mic IP'ing: http://dyndn.es
> !!! DynDN.eS is NOT dyn.com !!!
>
>

On 2016-01-14 19:51, Shepherd Nhongo wrote:
> Do you have glibc on your machine ?

It is glibc-2.21.

> On 14 Jan 2016 20:27, "K. Peter" <kp@dyndn.es> wrote:
>
>> Hi,
>>
>> I was coming back to an issue I saw in the past but didn't
>> anything. I think it is related to all qmail derivates. At least
>> qmail-1.03 and netqmail-1.06 are affected as s/qmail-3.01 too.
>>
>> During make several times gcc creates a warning for .c files which
>> uses the SUBSTDIO_FDBUF macro, defined in substdio.h. This example
>> reproduces it:
>>
>> # gcc -fsyntax-only subfderr.c
>> subfderr.c:8:1: warning: initialization from incompatible pointer
>> type
>> static substdio it = SUBSTDIO_FDBUF(write,2,subfd_errbuf,256);
>> ^
>> subfderr.c:8:1: warning: (near initialization for ‘it.op’)
>>
>> With my rudimental C expierience I would expect that a cast have to
>> be done, but maybe I'm wrong. So the questions are
>>
>> 1. Does anybody have had the same issue and did solve it? What is
>> the solution?
>> 2. Could anybody provide a solution or point me in the right
>> direction?
>> 3. In case a pointer cast is needed, where have it be done?
>>
>> System is an up-to-date Gentoo with gcc-4.9.3.
>>
>> Any help is appreciated.
>>
>> thanks and regards
>> Kai
>>
>> --
>>
> -⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠
>> Dyn@mic IP'ing: http://dyndn.es [1]
>> !!! DynDN.eS is NOT dyn.com [2] !!!
>
>
> Links:
> ------
> [1] http://dyndn.es
> [2] http://dyn.com

--
-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠
Dyn@mic IP'ing: http://dyndn.es
!!! DynDN.eS is NOT dyn.com !!!

On Thu, 14 Jan 2016 at 23:56 K. Peter <kp@dyndn.es> wrote:

> # gcc -fsyntax-only subfderr.c
> subfderr.c:8:1: warning: initialization from incompatible pointer type
> static substdio it = SUBSTDIO_FDBUF(write,2,subfd_errbuf,256);
> ^
> subfderr.c:8:1: warning: (near initialization for ‘it.op’)
>
>
> With my rudimental C expierience I would expect that a cast have to be
> done, but maybe I'm wrong. So the questions are
>
>
include unistd.h for the prototype for write() system call and abracadabra,
your warning will dissappear.

On 2016-01-15 04:38, Manvendra Bhangui wrote:
> On Thu, 14 Jan 2016 at 23:56 K. Peter <kp@dyndn.es> wrote:
>
>> # gcc -fsyntax-only subfderr.c
>> subfderr.c:8:1: warning: initialization from incompatible pointer
>> type
>> static substdio it = SUBSTDIO_FDBUF(write,2,subfd_errbuf,256);
>> ^
>> subfderr.c:8:1: warning: (near initialization for ‘it.op’)
>>
>> With my rudimental C expierience I would expect that a cast have to
>> be
>> done, but maybe I'm wrong. So the questions are
>
> include unistd.h for the prototype for write() system call and
> abracadabra, your warning will dissappear.

Hi Manvendra,

thanks for the response, it puts me to the right direction.

I solved the issue by creating a readwrite2.h files:

# cat readwrite.h
#ifndef READWRITE_H
#define READWRITE_H

#include <unistd.h>

#endif
----
# cat readwrite2.h
#ifndef READWRITE2_H
#define READWRITE2_H

extern int read();
extern int write();

#endif

Now I included one of these in the needed .c files. It is IMHO a bit
quick and dirty, maybe I will investigate to find a better solution.

For those who interested in the background: the "extern" declarations of
the original readwrite.h are conflicting with the declarations in
unistd.h. This (may be) affects other files too, e.g. timeoutconn.c -
using unistd.h here will result in a warning of unused results by gcc.

I have to appologize that I gave an incorrect information initially. The
netqmail-1.06 source code is absolutely ok related to this case (didn't
check it but assume qmail-1.03 too). It was a mistake from my side - I
did put unistd.h in the source code before. It looks like sqmail-3.01
still have this issue (@Erwin?).

Btw, thanks for help.
Kai

--
-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠-⁠
Dyn@mic IP'ing: http://dyndn.es
!!! DynDN.eS is NOT dyn.com !!!

Hi,

this problem has been solved within s/qmail (code refactoring).


http://fehcom.de/sqmail/doxygen/readwrite_8h_source.html



> Am 15.01.2016 um 12:06 schrieb K. Peter <kp@dyndn.es>:
>
> On 2016-01-15 04:38, Manvendra Bhangui wrote:
>> On Thu, 14 Jan 2016 at 23:56 K. Peter <kp@dyndn.es> wrote:
>>> # gcc -fsyntax-only subfderr.c
>>> subfderr.c:8:1: warning: initialization from incompatible pointer
>>> type
>>> static substdio it = SUBSTDIO_FDBUF(write,2,subfd_errbuf,256);
>>> ^
>>> subfderr.c:8:1: warning: (near initialization for ‘it.op’)
>>> With my rudimental C expierience I would expect that a cast have to
>>> be
>>> done, but maybe I'm wrong. So the questions are
>> include unistd.h for the prototype for write() system call and
>> abracadabra, your warning will dissappear.
>
> Hi Manvendra,
>
> thanks for the response, it puts me to the right direction.
>
> I solved the issue by creating a readwrite2.h files:
>
> # cat readwrite.h
> #ifndef READWRITE_H
> #define READWRITE_H
>
> #include <unistd.h>
>
> #endif
> ----
> # cat readwrite2.h
> #ifndef READWRITE2_H
> #define READWRITE2_H
>
> extern int read();
> extern int write();
>
> #endif
>
> Now I included one of these in the needed .c files. It is IMHO a bit quick and dirty, maybe I will investigate to find a better solution.
>
> For those who interested in the background: the "extern" declarations of the original readwrite.h are conflicting with the declarations in unistd.h. This (may be) affects other files too, e.g. timeoutconn.c - using unistd.h here will result in a warning of unused results by gcc.
>
> I have to appologize that I gave an incorrect information initially. The netqmail-1.06 source code is absolutely ok related to this case (didn't check it but assume qmail-1.03 too). It was a mistake from my side - I did put unistd.h in the source code before. It looks like sqmail-3.01 still have this issue (@Erwin?).
>
> Btw, thanks for help.
> Kai
>
>

Best regards.
--eh.

Dr. Erwin Hoffmann | FEHCom | http://www.fehcom.de <http://www.fehcom.de/> | PGP Key-Id: EE00CF6