Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Since there was the upgrade from GCC5 to GCC6 on MSYS2-MinGW64, the 
build log of Emacs (master) contains a lot of warnings like this:

[...]/work/emacs-master/nt/inc/ms-w32.h:296:17: warning: conflicting 
types for built-in function 'execve'

$ grep -c 'conflicting types' 
emacs-master-w64-x86_64-20160830_205442-release-build.log
134

Maybe you can silence them...


Thanks,
  Angelo.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Wed, 31 Aug 2016 10:01:25 +0200
> 
> Since there was the upgrade from GCC5 to GCC6 on MSYS2-MinGW64, the 
> build log of Emacs (master) contains a lot of warnings like this:
> 
> [...]/work/emacs-master/nt/inc/ms-w32.h:296:17: warning: conflicting 
> types for built-in function 'execve'

Please show at least one such warning in its entirety.  I don't have
GCC 6.x here, so I cannot know what does it not like in our prototype.

Thanks.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Il 31/08/2016 16:27, Eli Zaretskii ha scritto:
>> From: Angelo Graziosi <angelo.graziosi@alice.it>
>> Date: Wed, 31 Aug 2016 10:01:25 +0200
>>
>> Since there was the upgrade from GCC5 to GCC6 on MSYS2-MinGW64, the
>> build log of Emacs (master) contains a lot of warnings like this:
>>
>> [...]/work/emacs-master/nt/inc/ms-w32.h:296:17: warning: conflicting
>> types for built-in function 'execve'
>
> Please show at least one such warning in its entirety.  I don't have
> GCC 6.x here, so I cannot know what does it not like in our prototype.

Oh yes, this is an example

make[2]: ingresso nella directory "/tmp/work/emacs-master/lib"
   CC       acl-errno-valid.o
   CCLD     ddeclient.exe
In file included from ../src/conf_post.h:32:0,

                  from ../src/config.h:1926,

                  from acl-errno-valid.c:20:

C:/msys64/tmp/work/emacs-master/nt/inc/ms-w32.h:296:17: warning: 
conflicting types for built-in function 'execve'

  extern intptr_t execve (const char *, char * const *, char * const *);

                  ^~~~~~

   CCLD     runemacs.exe
make[1]: uscita dalla directory "/tmp/work/emacs-master/nt"



The header ms-w32.h is included elsewhere and gives 134 of this warnings..

> Thanks.

Angelo

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Cc: emacs-devel@gnu.org
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Wed, 31 Aug 2016 17:01:08 +0200
> 
> > Please show at least one such warning in its entirety.  I don't have
> > GCC 6.x here, so I cannot know what does it not like in our prototype.
> 
> Oh yes, this is an example
> 
> make[2]: ingresso nella directory "/tmp/work/emacs-master/lib"
>    CC       acl-errno-valid.o
>    CCLD     ddeclient.exe
> In file included from ../src/conf_post.h:32:0,
> 
>                   from ../src/config.h:1926,
> 
>                   from acl-errno-valid.c:20:
> 
> C:/msys64/tmp/work/emacs-master/nt/inc/ms-w32.h:296:17: warning: 
> conflicting types for built-in function 'execve'
> 
>   extern intptr_t execve (const char *, char * const *, char * const *);
> 
>                   ^~~~~~
> 

Sigh... so unhelpful.

If you change the prototype in ms-w32.h to say this:

  extern intptr_t execve (const char *, char * const [], char * const []);

do these warnings go away?

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Il 31/08/2016 17:42, Eli Zaretskii ha scritto:
>
> Sigh... so unhelpful.
>
> If you change the prototype in ms-w32.h to say this:
>
>   extern intptr_t execve (const char *, char * const [], char * const []);
>
> do these warnings go away?
>

Really I already did this with this simple test case:

$ cat test.c
#include <stdint.h>

/*extern intptr_t execve (const char *, char * const *, char * const *);*/
extern intptr_t execve (const char *, char * const [], char * const []);

int main()
{
   return 0;
}

$ gcc test.c
test.c:4:17: warning: conflicting types for built-in function ‘execve’
  extern intptr_t execve (const char *, char * const [], char * const []);


but always the warning... I doubt things change modifying ms-w32.h..

Angelo

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Ken Brown]

On 8/31/2016 1:40 PM, Angelo Graziosi wrote:
> $ cat test.c
> #include <stdint.h>
>
> /*extern intptr_t execve (const char *, char * const *, char * const *);*/
> extern intptr_t execve (const char *, char * const [], char * const []);
>
> int main()
> {
>   return 0;
> }
>
> $ gcc test.c
> test.c:4:17: warning: conflicting types for built-in function ‘execve’
>  extern intptr_t execve (const char *, char * const [], char * const []);

According to

   http://pubs.opengroup.org/onlinepubs/9699919799/functions/exec.html
   http://man7.org/linux/man-pages/man2/execve.2.html

you should include <unistd.h>, and the return type should be int.

Ken

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Cc: emacs-devel@gnu.org
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Wed, 31 Aug 2016 19:40:50 +0200
> 
> > If you change the prototype in ms-w32.h to say this:
> >
> >   extern intptr_t execve (const char *, char * const [], char * const []);
> >
> > do these warnings go away?
> >
> 
> Really I already did this with this simple test case:

I couldn't know you did that already, could I?

> $ cat test.c
> #include <stdint.h>
> 
> /*extern intptr_t execve (const char *, char * const *, char * const *);*/
> extern intptr_t execve (const char *, char * const [], char * const []);
> 
> int main()
> {
>    return 0;
> }
> 
> $ gcc test.c
> test.c:4:17: warning: conflicting types for built-in function ‘execve’
>   extern intptr_t execve (const char *, char * const [], char * const []);

Then I don't know what to do about this.  Since GCC doesn't show the
prototype of its built-in, I have no idea which part of the prototype
triggers the warning.  The above suggestion was taken straight out of
the MinGW64 headers, which is the best I can do.

Maybe ask on the MinGW64 forum about this.  They might be more privy
to GCC internals than I am.

Sorry.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Cc: emacs-devel@gnu.org
> From: Ken Brown <kbrown@cornell.edu>
> Date: Wed, 31 Aug 2016 14:11:41 -0400
> 
> According to
> 
>    http://pubs.opengroup.org/onlinepubs/9699919799/functions/exec.html
>    http://man7.org/linux/man-pages/man2/execve.2.html
> 
> you should include <unistd.h>, and the return type should be int.

Thanks.  On MS-Windows, execve is not declared in unistd.h, it's
declared in process.h (which we cannot include, because it conflicts
with our own process.h).  Nevertheless, doing what you suggest is
worth a try, since there are no other ideas.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Il 31/08/2016 20:47, Eli Zaretskii ha scritto:
>> Cc: emacs-devel@gnu.org
>> From: Ken Brown <kbrown@cornell.edu>
>> Date: Wed, 31 Aug 2016 14:11:41 -0400
>>
>> According to
>>
>>    http://pubs.opengroup.org/onlinepubs/9699919799/functions/exec.html
>>    http://man7.org/linux/man-pages/man2/execve.2.html
>>
>> you should include <unistd.h>, and the return type should be int.
>
> Thanks.  On MS-Windows, execve is not declared in unistd.h, it's
> declared in process.h (which we cannot include, because it conflicts
> with our own process.h).  Nevertheless, doing what you suggest is
> worth a try, since there are no other ideas.
>

As always, Ken's suggestions are right!

Now this test case,

$ cat test.c
/*#include <stdint.h>*/
#include <unistd.h>

/*extern intptr_t execve (const char *, char * const *, char * const *);*/
/*extern intptr_t execve (const char *, char * const [], char * const []);*/
extern int execve (const char *, char * const *, char * const *);

int main()
{
   return 0;
}

builds without warnings with

$ gcc -Wall test.c


BTW, I tried my first test case (that with the Emacs definition of 
'execve') with the other compilers in MSYS2/MinGW-64/32, and both MSYS2 
gcc 5.3 and MinGW64 gcc-6.1 give the warning, instead MinGW32 gcc-6.1 
does not print warnings.

Ciao,
Angelo.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Wed, 31 Aug 2016 21:56:45 +0200
> Cc: emacs-devel@gnu.org
> 
> As always, Ken's suggestions are right!
> 
> Now this test case,
> 
> $ cat test.c
> /*#include <stdint.h>*/
> #include <unistd.h>
> 
> /*extern intptr_t execve (const char *, char * const *, char * const *);*/
> /*extern intptr_t execve (const char *, char * const [], char * const []);*/
> extern int execve (const char *, char * const *, char * const *);
> 
> int main()
> {
>    return 0;
> }
> 
> builds without warnings with
> 
> $ gcc -Wall test.c

Thanks.  What happens if you remove the unistd.h inclusion?

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Il 01/09/2016 04:37, Eli Zaretskii ha scritto:
>> From: Angelo Graziosi <angelo.graziosi@alice.it>
>> Date: Wed, 31 Aug 2016 21:56:45 +0200
>> Cc: emacs-devel@gnu.org
>>
>> As always, Ken's suggestions are right!
>>
>> Now this test case,
>>
>> $ cat test.c
>> /*#include <stdint.h>*/
>> #include <unistd.h>
>>
>> /*extern intptr_t execve (const char *, char * const *, char * const *);*/
>> /*extern intptr_t execve (const char *, char * const [], char * const []);*/
>> extern int execve (const char *, char * const *, char * const *);
>>
>> int main()
>> {
>>    return 0;
>> }
>>
>> builds without warnings with
>>
>> $ gcc -Wall test.c
>
> Thanks.  What happens if you remove the unistd.h inclusion?
>

It builds without warnings. So the simplest test case is:


extern int execve (const char *, char * const *, char * const *);

int main()
{
   return 0;
}


Ciao, Angelo.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Cc: kbrown@cornell.edu, emacs-devel@gnu.org
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Thu, 1 Sep 2016 09:44:59 +0200
> 
> > Thanks.  What happens if you remove the unistd.h inclusion?
> >
> 
> It builds without warnings. So the simplest test case is:
> 
> 
> extern int execve (const char *, char * const *, char * const *);
> 
> int main()
> {
>    return 0;
> }

Thanks, I made the changes on the master branch that are supposed to
fix the warnings.  Please test.

The prototype that fixes this, with 'int' as the return value, is
wrong, and is IMO a MinGW64 GCC bug which should be reported.
intptr_t is a 64-bit type in MinGW64, while int is still 32 bits.  By
enforcing the wrong type, GCC contradicts MinGW64's own library
headers.

Btw, another possibility to avoid the warnings would be to use the GCC
option -fno-builtin-execve.  Can you try that (with the previous
prototype)?

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Il 01/09/2016 16:17, Eli Zaretskii ha scritto:
>
> The prototype that fixes this, with 'int' as the return value, is
> wrong, and is IMO a MinGW64 GCC bug which should be reported.
> intptr_t is a 64-bit type in MinGW64, while int is still 32 bits.  By
> enforcing the wrong type, GCC contradicts MinGW64's own library
> headers.

I tested the test case (that using intptr_t) on OSX with gcc-6.1 and it 
prints the same warning only using 'int' as return type the warning goes 
away...

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Cc: kbrown@cornell.edu, emacs-devel@gnu.org
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Thu, 1 Sep 2016 17:42:06 +0200
> 
> > The prototype that fixes this, with 'int' as the return value, is
> > wrong, and is IMO a MinGW64 GCC bug which should be reported.
> > intptr_t is a 64-bit type in MinGW64, while int is still 32 bits.  By
> > enforcing the wrong type, GCC contradicts MinGW64's own library
> > headers.
> 
> I tested the test case (that using intptr_t) on OSX with gcc-6.1 and it 
> prints the same warning only using 'int' as return type the warning goes 
> away...

On Unix, execve indeed returns an int, because it's a process ID.  On
Windows, the return value can be a process handle, which is a 64-bit
data type on 64-bit Windows.  That is why the return value must be
intptr_t, not an int.

The bug in MinGW64 GCC is that it uses the Unix model for the builtin
execve, whereas it should have used the Windows model.

Anyway, if you don't believe me, look on MSDN:

  https://msdn.microsoft.com/en-us/library/hyw61wtd.aspx

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Date: Thu, 01 Sep 2016 19:04:45 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: kbrown@cornell.edu, emacs-devel@gnu.org
> 
> On Unix, execve indeed returns an int, because it's a process ID.

Sorry, that was wrong: it's an int error code, not a PID.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Il 01/09/2016 18:04, Eli Zaretskii ha scritto:
>
> Anyway, if you don't believe me, look on MSDN:
>
>   https://msdn.microsoft.com/en-us/library/hyw61wtd.aspx
>

I haven't reasons to doubt...

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Davis Herring]

> On Unix, execve indeed returns an int, because it's a process ID.  On
> Windows, the return value can be a process handle, which is a 64-bit
> data type on 64-bit Windows.  That is why the return value must be
> intptr_t, not an int.

No, it's an error code.  (You said as much for POSIX in your following 
message, but it's an error code on all systems.)  If it succeeds, the 
process image has been replaced and it doesn't return at all.

Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or 
too sparse, it is because mass-energy conversion has occurred during 
shipping.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Ken Brown]

On 9/1/2016 12:17 PM, Davis Herring wrote:
>> On Unix, execve indeed returns an int, because it's a process ID.  On
>> Windows, the return value can be a process handle, which is a 64-bit
>> data type on 64-bit Windows.  That is why the return value must be
>> intptr_t, not an int.
>
> No, it's an error code.  (You said as much for POSIX in your following
> message, but it's an error code on all systems.)  If it succeeds, the
> process image has been replaced and it doesn't return at all.

The return value is not an error code.  If the function returns, it 
returns -1 and sets errno:

   http://pubs.opengroup.org/onlinepubs/9699919799/functions/exec.html
   https://msdn.microsoft.com/en-us/library/hyw61wtd.aspx

Ken

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Il 01/09/2016 16:17, Eli Zaretskii ha scritto:
>
> Btw, another possibility to avoid the warnings would be to use the GCC
> option -fno-builtin-execve.  Can you try that (with the previous
> prototype)?
>

I tested with the test case and it works: no warnings.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Cc: Angelo Graziosi <angelo.graziosi@alice.it>, kbrown@cornell.edu,
>         emacs-devel@gnu.org
> From: Davis Herring <herring@lanl.gov>
> Date: Thu, 1 Sep 2016 10:17:36 -0600
> 
> > On Unix, execve indeed returns an int, because it's a process ID.  On
> > Windows, the return value can be a process handle, which is a 64-bit
> > data type on 64-bit Windows.  That is why the return value must be
> > intptr_t, not an int.
> 
> No, it's an error code.  (You said as much for POSIX in your following 
> message, but it's an error code on all systems.)  If it succeeds, the 
> process image has been replaced and it doesn't return at all.

On Windows, the error code is returned from a function that can also
return a process handle.  Look in the MS CRT sources, and you will see
it.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Cc: kbrown@cornell.edu, emacs-devel@gnu.org
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Date: Thu, 1 Sep 2016 18:25:34 +0200
> 
> Il 01/09/2016 16:17, Eli Zaretskii ha scritto:
> >
> > Btw, another possibility to avoid the warnings would be to use the GCC
> > option -fno-builtin-execve.  Can you try that (with the previous
> > prototype)?
> >
> 
> I tested with the test case and it works: no warnings.

Thanks.  Not sure if we should use that alternative.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Andreas Schwab]

On Sep 01 2016, Eli Zaretskii <eliz@gnu.org> wrote:

> Anyway, if you don't believe me, look on MSDN:
>
>   https://msdn.microsoft.com/en-us/library/hyw61wtd.aspx

execve != _execve.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> From: Andreas Schwab <schwab@linux-m68k.org>
> Cc: Angelo Graziosi <angelo.graziosi@alice.it>,  kbrown@cornell.edu,  emacs-devel@gnu.org
> Date: Thu, 01 Sep 2016 18:32:44 +0200
> 
> On Sep 01 2016, Eli Zaretskii <eliz@gnu.org> wrote:
> 
> > Anyway, if you don't believe me, look on MSDN:
> >
> >   https://msdn.microsoft.com/en-us/library/hyw61wtd.aspx
> 
> execve != _execve.

It is in MinGW.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Eli Zaretskii]

> Date: Thu, 01 Sep 2016 17:17:35 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: kbrown@cornell.edu, emacs-devel@gnu.org
> 
> The prototype that fixes this, with 'int' as the return value, is
> wrong, and is IMO a MinGW64 GCC bug which should be reported.
> intptr_t is a 64-bit type in MinGW64, while int is still 32 bits.  By
> enforcing the wrong type, GCC contradicts MinGW64's own library
> headers.

Actually, I see now that the prototype of execve in MinGW64 library
headers does match the builtin, and that of _execve doesn't (although
if the library function execve is called, it will call _execve).  So I
guess it's not a bug, strictly speaking, but a kludgey misfeature, and
a deviation from mingw.org's MinGW headers.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Davis Herring]

> The return value is not an error code.  If the function returns, it
> returns -1 and sets errno:

Sorry -- I borrowed the phrase from Eli's subsequent email, and (I 
think) we both meant "a return value indicating whether an error 
occurred".  For execve(), of course, it is simply "yes".

Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or 
too sparse, it is because mass-energy conversion has occurred during 
shipping.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > > Really I already did this with this simple test case:

  > I couldn't know you did that already, could I?

It would have been better to say,

  That was good.  Too bad it didn't resolve the problem
  as we hoped it would.

I think you interpreted "Really I already did this with this simple
test case" as a rebuke to you: "Don't criticize me for not trying that
test case!  I DID try it!"

So you defended yourself by saying that you couldn't possibly
have known he had tried it.

That defense is valid -- but instead of defending yourself, it is much
better to praise/thank him for doing the right thing.

It will also defend you, because people will see you did not know he
had not tried that test.  Indeed, it will defend you _better_ than
what you actually said, because people will admire you for that
positivity and constructiveness.

I have had to learn to do this, over the years, and I have not learned
100%.  I still don't always manage to do it.  But I often do, and that
helps me get better outcomes.  You can learn too.

-- 
Dr Richard Stallman
President, Free Software Foundation (gnu.org, fsf.org)
Internet Hall-of-Famer (internethalloffame.org)
Skype: No way! See stallman.org/skype.html.

Re: Too many warnings building Emacs with GCC 6 on MSYS2-MinGW64

[Angelo Graziosi]

Il 01/09/2016 18:30, Eli Zaretskii ha scritto:
>> Cc: kbrown@cornell.edu, emacs-devel@gnu.org
>> From: Angelo Graziosi <angelo.graziosi@alice.it>
>> Date: Thu, 1 Sep 2016 18:25:34 +0200
>>
>> Il 01/09/2016 16:17, Eli Zaretskii ha scritto:
>>>
>>> Btw, another possibility to avoid the warnings would be to use the GCC
>>> option -fno-builtin-execve.  Can you try that (with the previous
>>> prototype)?
>>>
>>
>> I tested with the test case and it works: no warnings.
>
> Thanks.  Not sure if we should use that alternative.
>

+1

Angelo