mingw-users

[Mingw-users] How to use the fseeko64() function ?

[Julien Darthenay <julie...@la...>]

Hi everybody,
I had a problem reading a big file (more then 2 gigabytes) because I 
need to use fseek() and ftell() to do what I want with it. I've been 
told about the long long data type which I didn't know it exists and the 
fseeko() and ftello() functions. I've seen there is no fseeko() function 
in my stdio.h, but there is a fseeko64() function and a ftello64() 
function which should be what I need. But their definitions are 
optionnal and depend of my  #defines. Which #define should I add in my 
source code (or -D option in my makefile) in order to use these 
functions without any warning ?

My Mingw runtime is : mingwrt-3.15.1-mingw32.tar.gz
My OS is Windows XP SP2.

Thanks for your help.

Julien Darthenay

Re: [Mingw-users] How to use the fseeko64() function ?

[Greg Chicares <gchic...@sb...>]

On 2009-03-24 22:18Z, Julien Darthenay wrote:
>
> I had a problem reading a big file (more then 2 gigabytes) because I 
> need to use fseek() and ftell() to do what I want with it. I've been 
> told about the long long data type which I didn't know it exists and the 
> fseeko() and ftello() functions. I've seen there is no fseeko() function 
> in my stdio.h, but there is a fseeko64() function and a ftello64() 
> function which should be what I need. But their definitions are 
> optionnal and depend of my  #defines. Which #define should I add in my 
> source code (or -D option in my makefile) in order to use these 
> functions without any warning ?

Here's an fseeko64() testcase that is said to need no '-D' option:
  http://article.gmane.org/gmane.comp.gnu.mingw.user/28881
Does it work for you? If that doesn't lead to an answer, then post
a minimal standalone testcase of your own, copying and pasting any
diagnostics you observe when compiling it.

Re: [Mingw-users] How to use the fseeko64() function ?

[Julien Darthenay <julie...@la...>]

Greg Chicares a écrit :
> On 2009-03-24 22:18Z, Julien Darthenay wrote:
>   
>> I had a problem reading a big file (more then 2 gigabytes) because I 
>> need to use fseek() and ftell() to do what I want with it. I've been 
>> told about the long long data type which I didn't know it exists and the 
>> fseeko() and ftello() functions. I've seen there is no fseeko() function 
>> in my stdio.h, but there is a fseeko64() function and a ftello64() 
>> function which should be what I need. But their definitions are 
>> optionnal and depend of my  #defines. Which #define should I add in my 
>> source code (or -D option in my makefile) in order to use these 
>> functions without any warning ?
>>     
>
> Here's an fseeko64() testcase that is said to need no '-D' option:
>   http://article.gmane.org/gmane.comp.gnu.mingw.user/28881
> Does it work for you? If that doesn't lead to an answer, then post
> a minimal standalone testcase of your own, copying and pasting any
> diagnostics you observe when compiling it.
>
> ------------------------------------------------------------------------------
> Apps built with the Adobe(R) Flex(R) framework and Flex Builder(TM) are
> powering Web 2.0 with engaging, cross-platform capabilities. Quickly and
> easily build your RIAs with Flex Builder, the Eclipse(TM)based development
> software that enables intelligent coding and step-through debugging.
> Download the free 60 day trial. http://p.sf.net/sfu/www-adobe-com
>   
This don't work. I get the following warnings : (I created the 
fseeko64.c file to test this sample code)
fseeko64.c:15: warning: implicit declaration of function `fileno'
fseeko64.c:22: warning: implicit declaration of function `fseeko64'
fseeko64.c:23: warning: implicit declaration of function `ftello64'


You can't reasonably use a function with a warning and expect it to work.

Re: [Mingw-users] How to use the fseeko64() function ?

[Greg Chicares <gchic...@sb...>]

On 2009-03-25 22:36Z, Julien Darthenay wrote:
> Greg Chicares a écrit :
>> On 2009-03-24 22:18Z, Julien Darthenay wrote:
>>   
>>> I had a problem reading a big file (more then 2 gigabytes) because I 
>>> need to use fseek() and ftell() to do what I want with it. I've been 
>>> told about the long long data type which I didn't know it exists and the 
>>> fseeko() and ftello() functions. I've seen there is no fseeko() function 
>>> in my stdio.h, but there is a fseeko64() function and a ftello64() 
>>> function which should be what I need. But their definitions are 
>>> optionnal and depend of my  #defines. Which #define should I add in my 
>>> source code (or -D option in my makefile) in order to use these 
>>> functions without any warning ?
>>
>> Here's an fseeko64() testcase that is said to need no '-D' option:
>>   http://article.gmane.org/gmane.comp.gnu.mingw.user/28881
[...]
> This don't work. I get the following warnings : (I created the 
> fseeko64.c file to test this sample code)
> fseeko64.c:15: warning: implicit declaration of function `fileno'
> fseeko64.c:22: warning: implicit declaration of function `fseeko64'
> fseeko64.c:23: warning: implicit declaration of function `ftello64'

I copied the testcase directly from Tor's message, and I didn't
get those warnings. Here's what I see with fairly finicky warning
options, expressly including '-Wimplicit':

$/MinGW_/bin/gcc -Wall -W -pedantic -Wimplicit -c fseeko64.c
fseeko64.c:22:16: warning: use of C99 long long integer constant
fseeko64.c: In function `main':
fseeko64.c:23: warning: ISO C does not support the `I64' printf length modifier
fseeko64.c:25: warning: ISO C does not support the `I64' printf length modifier
fseeko64.c:32:16: warning: use of C99 long long integer constant
fseeko64.c:33: warning: ISO C does not support the `I64' printf length modifier
fseeko64.c: At top level:
fseeko64.c:9: warning: unused parameter 'argc'
fseeko64.c:9: warning: unused parameter 'argv'

That's with MinGW gcc-3.4.4, but I also tried 3.4.5 and 4.3.0,
and didn't see any of the warnings you reported. Are you using
different compiler options?

Re: [Mingw-users] How to use the fseeko64() function ?

[Julien Darthenay <julie...@la...>]

Greg Chicares a écrit :
> On 2009-03-25 22:36Z, Julien Darthenay wrote:
>   
>> Greg Chicares a écrit :
>>     
>>> On 2009-03-24 22:18Z, Julien Darthenay wrote:
>>>   
>>>       
>>>> I had a problem reading a big file (more then 2 gigabytes) because I 
>>>> need to use fseek() and ftell() to do what I want with it. I've been 
>>>> told about the long long data type which I didn't know it exists and the 
>>>> fseeko() and ftello() functions. I've seen there is no fseeko() function 
>>>> in my stdio.h, but there is a fseeko64() function and a ftello64() 
>>>> function which should be what I need. But their definitions are 
>>>> optionnal and depend of my  #defines. Which #define should I add in my 
>>>> source code (or -D option in my makefile) in order to use these 
>>>> functions without any warning ?
>>>>         
>>> Here's an fseeko64() testcase that is said to need no '-D' option:
>>>   http://article.gmane.org/gmane.comp.gnu.mingw.user/28881
>>>       
> [...]
>   
>> This don't work. I get the following warnings : (I created the 
>> fseeko64.c file to test this sample code)
>> fseeko64.c:15: warning: implicit declaration of function `fileno'
>> fseeko64.c:22: warning: implicit declaration of function `fseeko64'
>> fseeko64.c:23: warning: implicit declaration of function `ftello64'
>>     
>
> I copied the testcase directly from Tor's message, and I didn't
> get those warnings. Here's what I see with fairly finicky warning
> options, expressly including '-Wimplicit':
>
> $/MinGW_/bin/gcc -Wall -W -pedantic -Wimplicit -c fseeko64.c
> fseeko64.c:22:16: warning: use of C99 long long integer constant
> fseeko64.c: In function `main':
> fseeko64.c:23: warning: ISO C does not support the `I64' printf length modifier
> fseeko64.c:25: warning: ISO C does not support the `I64' printf length modifier
> fseeko64.c:32:16: warning: use of C99 long long integer constant
> fseeko64.c:33: warning: ISO C does not support the `I64' printf length modifier
> fseeko64.c: At top level:
> fseeko64.c:9: warning: unused parameter 'argc'
> fseeko64.c:9: warning: unused parameter 'argv'
>
> That's with MinGW gcc-3.4.4, but I also tried 3.4.5 and 4.3.0,
> and didn't see any of the warnings you reported. Are you using
> different compiler options?
>
> ------------------------------------------------------------------------------
>   
Hi,

I've just discovered a way to solve the warnings problem. I added the 
following directive before including <stdio.h> :
#undef __STRICT_ANSI__

I don't know why __STRICT_ANSI__ is defined even when I don't use the 
-ansi option. My makefile contains the following flags :
CFLAGS = -Wformat -Wall -std=c99
LDFLAGS = -lcomctl32
UNICODEFLAGS = -DUNICODE -D_UNICODE -finput-charset=iso-8859-1

Re: [Mingw-users] How to use the fseeko64() function ?

[Greg Chicares <gchic...@sb...>]

On 2009-03-26 11:40Z, Julien Darthenay wrote:
> 
> I've just discovered a way to solve the warnings problem. I added the 
> following directive before including <stdio.h> :
> #undef __STRICT_ANSI__
> 
> I don't know why __STRICT_ANSI__ is defined even when I don't use the 
> -ansi option. My makefile contains the following flags :
> CFLAGS = -Wformat -Wall -std=c99

You have specified the '-std=c99' option, but you're using
functions like fseeko64() that are not part of the C99 standard,
so you get a warning. (You could say that __STRICT_ANSI__ is a
misnomer: the compiler sets it for all '-std' options that
specify any ISO standard.)

That's actually the correct behavior. Consider:

   #include <stdio.h>
   void foo(void);
   int main() {foo(); return 0;}

That's a strictly-conforming C99 program if you replace 'foo'
with any name that C99 does not reserve, including 'fseeko64',
so 'gcc -std=c99' must accept that program.

Of course, that behavior is inconvenient when you want to use
extensions like fseeko64(). Options, from worst to best IMO:
 (1) Don't use '-std='. But then you may not get other valuable
     diagnostics that you really do want.
 (2) Prototype all extra functions yourself.
 (3) Manipulate __STRICT_ANSI__.
 (4) Use a different '-std=' setting if possible.
In this case, '-std=gnu99' works, and that's probably the best
option. (Where it doesn't work, I prefer (3) myself.)

Re: [Mingw-users] How to use the fseeko64() function ?

[Julien Darthenay <julie...@la...>]

Greg Chicares a écrit :
> On 2009-03-26 11:40Z, Julien Darthenay wrote:
>   
>> I've just discovered a way to solve the warnings problem. I added the 
>> following directive before including <stdio.h> :
>> #undef __STRICT_ANSI__
>>
>> I don't know why __STRICT_ANSI__ is defined even when I don't use the 
>> -ansi option. My makefile contains the following flags :
>> CFLAGS = -Wformat -Wall -std=c99
>>     
>
> You have specified the '-std=c99' option, but you're using
> functions like fseeko64() that are not part of the C99 standard,
> so you get a warning. (You could say that __STRICT_ANSI__ is a
> misnomer: the compiler sets it for all '-std' options that
> specify any ISO standard.)
>
> That's actually the correct behavior. Consider:
>
>    #include <stdio.h>
>    void foo(void);
>    int main() {foo(); return 0;}
>
> That's a strictly-conforming C99 program if you replace 'foo'
> with any name that C99 does not reserve, including 'fseeko64',
> so 'gcc -std=c99' must accept that program.
>
> Of course, that behavior is inconvenient when you want to use
> extensions like fseeko64(). Options, from worst to best IMO:
>  (1) Don't use '-std='. But then you may not get other valuable
>      diagnostics that you really do want.
>  (2) Prototype all extra functions yourself.
>  (3) Manipulate __STRICT_ANSI__.
>  (4) Use a different '-std=' setting if possible.
> In this case, '-std=gnu99' works, and that's probably the best
> option. (Where it doesn't work, I prefer (3) myself.)
>
> ------------------------------------------------------------------------------
>   
Ok, but I had to add -std=c99 to have the long long type recognized, 
wich is necessary to use fseeko64(). I tried with -std=gnu99 and it 
worked perfectly, than you very much.

  Julien Darthenay

Re: [Mingw-users] How to use the fseeko64() function ?

[Keith Marshall <keith...@us...>]

On Wednesday 25 March 2009 22:36:10 Julien Darthenay wrote:
> > Here's an fseeko64() testcase that is said to need no '-D'
> > option: http://article.gmane.org/gmane.comp.gnu.mingw.user/28881
> > Does it work for you? If that doesn't lead to an answer, then
> > post a minimal standalone testcase of your own, copying and
> > pasting any diagnostics you observe when compiling it.
>
> This don't work.

It does for me.

> I get the following warnings : (I created the 
> fseeko64.c file to test this sample code)
> fseeko64.c:15: warning: implicit declaration of function `fileno'
> fseeko64.c:22: warning: implicit declaration of function `fseeko64'
> fseeko64.c:23: warning: implicit declaration of function `ftello64'

You must have compiled with `-ansi', or one of the `-std=...' options 
which implies `__STRICT_ANSI__'.  None of the three functions above are 
required by ISO C99 (ANSI), and their prototypes are not exposed when you 
demand strict standards compliance.

> You can't reasonably use a function with a warning and expect it to
> work.

You can, if you understand the warning sufficiently to deem it to be 
irrelevant; for example, I see much the same as Greg:

$ gcc -pedantic -Wall -Wextra -c foo.c
foo.c:21:16: warning: use of C99 long long integer constant
foo.c: In function `main':
foo.c:22: warning: ISO C does not support the `I64' printf length modifier
foo.c:25: warning: ISO C does not support the `I64' printf length modifier
foo.c:32:16: warning: use of C99 long long integer constant
foo.c:33: warning: ISO C does not support the `I64' printf length modifier
foo.c: At top level:
foo.c:9: warning: unused parameter 'argc'
foo.c:9: warning: unused parameter 'argv'

Here we can safely ignore all of those warnings, because:

1) The C99 long long integer constant is passed as an argument to a
   function, where the prototype expects a `typedef long long'.
2) Our specific printf() implementation *does* support the `I64'
   length modifier.
3) It hardly matters that we never use the arguments to main(); (of
   course, it would be better to define it without any).

That said, implicit declaration of a function is a warning which you 
should not simply ignore; you should provide a prototype, and in this 
case you can do that by *not* excluding non-ansi functions.

-- 

Regards,
Keith.