[ECOS] [Fwd: [ECOS] printf and floating point emulation]

[ECOS] [Fwd: [ECOS] printf and floating point emulation]

[Eduardo Sabaj]

I found that the problem is the function vfnprint () located in 
vfnprintf.cxx. It calls cvt() and this function calls modf(). This last 
function seems to be the problem. This is the code extracted from cvt() :
                ...
        for (p = endp - 1; integer; ++expcnt) {
                tmp = modf(integer / 10, &integer);
                *p-- = to_char((int)((tmp + .01) * 10));
        }
                ...

This loop never ends!

Am I omiting any patch? Please I need help. If you have any suggestion 
write me as soon as you can.
Regards,
Eduardo


-------- Mensaje original --------
Asunto: 	[ECOS] printf and floating point emulation
Fecha: 	Tue, 08 Apr 2008 16:55:09 -0300
De: 	Eduardo Sabaj <esabaj@hasar.com>
Para: 	ecos-discuss@ecos.sourceware.org



I have a program that hangs when trying to format and print something 
like this:

void
func (float f)
{
   printf ("%.2f\n", f);
}

I don't have an FPU builtin hardware.
The CPPFLAGS and LDFLAGS include the following options:

-msoft-float -mno-fp-ret-in-387 -mno-fancy-math-387

I disabled the FPU support (macro CYGHWR_HAL_I386_FPU) and I enabled 
CYGSEM_LIBC_STDIO_PRINTF_FLOATING_POINT and 
CYGSEM_LIBC_STDIO_SCANF_FLOATING_POINT macros.

I'm running ecos-2.0 (the last version available in CVS).

Please, I cannot realize what I'm omiting.. Any idea?

-- 

Ing. Eduardo Salvador Sabaj
Analista de Desarrollo

Compañía Hasar | Grupo Hasar*
*Marcos Sastre y José Ingenieros
El Talar. Pacheco
[B1618CSD] Buenos Aires. Argentina
Tel [54 11] 4117 8900 | Fax [54 11] 4117 8998
E-mail: esabaj@hasar.com
Visítenos en: www.hasar.com
<http://www.hasar.com>Información legal y política de confidencialidad: 
www.grupohasar.com/disclaimer <http://www.grupohasar.com/disclaimer>


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss




<http://www.grupohasar.com/disclaimer> 


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] [Fwd: [ECOS] printf and floating point emulation]

[Andrew Lunn]

On Wed, Apr 09, 2008 at 01:04:03PM -0300, Eduardo Sabaj wrote:
> I found that the problem is the function vfnprint () located in  
> vfnprintf.cxx. It calls cvt() and this function calls modf(). This last  
> function seems to be the problem. This is the code extracted from cvt() :
>                ...
>        for (p = endp - 1; integer; ++expcnt) {
>                tmp = modf(integer / 10, &integer);
>                *p-- = to_char((int)((tmp + .01) * 10));
>        }
>                ...
>
> This loop never ends!
>
> Am I omiting any patch? Please I need help. If you have any suggestion  
> write me as soon as you can.
> Regards,
> Eduardo

Can you give a concrete example. Something like

main()
{
        printf("%.2f\n, 3.141539);
}

which causes the problem. Does the value being printed matter?

What happens when you run the test program
packages/language/c/libc/stdio/current/tests/sprintf1.c

      Thanks
        Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] [Fwd: [ECOS] printf and floating point emulation]

[Eduardo Sabaj]

Andrew, the program fails with the printing of any floating point 
number. Anyway, what I'm printing is :

        printf ("main: %f\n", 2.345);

The lines below were produced by the test program in:

	packages/language/c/libc/stdio/current/tests/sprintf1.c

When the program reaches the first line with a %f format it hangs:

	// Check 18
	ret = sprintf(x, "%f", 2.5);

By the way, I downloaded the latest eCos version with the following command:
	
	cvs -z3 -d :pserver:anoncvs@ecos.sourceware.org:/cvs/ecos co -P ecos

Any idea?

Regards,
Eduardo

----

INFO:<Starting tests from testcase sprintf1.c for C library sprintf() 
function>
INFO:<These test individual features separately>
PASS:<%d test>
PASS:<%d test return code>
PASS:<%s test>
PASS:<%s test return code>
PASS:<padding test>
PASS:<padding test return code>
PASS:<hex conversion (lowercase)>
PASS:<hex conv (lowercase) return code>
PASS:<hex conversion (uppercase)>
PASS:<hex conv (upperbase ) return code>
PASS:<%c test>
PASS:<%c test return code>
PASS:<octal conversion>
PASS:<octal conversion return code>
PASS:<%u test>
PASS:<%u test return code>
PASS:<%n test>
PASS:<%n test return code>
PASS:<%% test>
PASS:<%% test return code>
PASS:<%ld test>
PASS:<%ld test return code>
PASS:<%lu test>
PASS:<%lu test return code>
PASS:<%x test>
PASS:<%x test return code>
PASS:<%X test>
PASS:<%x test return code>
PASS:<%0x test>
PASS:<%0x test return code>
PASS:<%0X test>
PASS:<%0X test return code>
PASS:<%p test return code>
INFO:<Starting floating point specific tests>

----

Andrew Lunn escribió:
> On Wed, Apr 09, 2008 at 01:04:03PM -0300, Eduardo Sabaj wrote:
>   
>> I found that the problem is the function vfnprint () located in  
>> vfnprintf.cxx. It calls cvt() and this function calls modf(). This last  
>> function seems to be the problem. This is the code extracted from cvt() :
>>                ...
>>        for (p = endp - 1; integer; ++expcnt) {
>>                tmp = modf(integer / 10, &integer);
>>                *p-- = to_char((int)((tmp + .01) * 10));
>>        }
>>                ...
>>
>> This loop never ends!
>>
>> Am I omiting any patch? Please I need help. If you have any suggestion  
>> write me as soon as you can.
>> Regards,
>> Eduardo
>>     
>
> Can you give a concrete example. Something like
>
> main()
> {
>         printf("%.2f\n, 3.141539);
> }
>
> which causes the problem. Does the value being printed matter?
>
> What happens when you run the test program
> packages/language/c/libc/stdio/current/tests/sprintf1.c
>
>       Thanks
>         Andrew
>
>   


-- 

Ing. Eduardo Salvador Sabaj
Analista de Desarrollo

Compañía Hasar | Grupo Hasar*
*Marcos Sastre y José Ingenieros
El Talar. Pacheco
[B1618CSD] Buenos Aires. Argentina
Tel [54 11] 4117 8900 | Fax [54 11] 4117 8998
E-mail: esabaj@hasar.com
Visítenos en: www.hasar.com
<http://www.hasar.com>Información legal y política de confidencialidad: 
www.grupohasar.com/disclaimer <http://www.grupohasar.com/disclaimer>


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] [Fwd: [ECOS] printf and floating point emulation]

[Andrew Lunn]

On Wed, Apr 09, 2008 at 03:17:42PM -0300, Eduardo Sabaj wrote:
> Andrew, the program fails with the printing of any floating point  
> number. Anyway, what I'm printing is :
>
>        printf ("main: %f\n", 2.345);
>
> The lines below were produced by the test program in:
>
> 	packages/language/c/libc/stdio/current/tests/sprintf1.c
>
> When the program reaches the first line with a %f format it hangs:
>
> 	// Check 18
> 	ret = sprintf(x, "%f", 2.5);

What toolchain are you using? Is it the eCos one from
http://ecos.sourceware.org/getstart.html

     Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] [Fwd: [ECOS] printf and floating point emulation]

[Eduardo Sabaj]

I'm working with one of the pre-built toolchains for Intel x386 which 
I've downloaded some years ago. Anyway, yesterday I downloaded this same 
toolchain from  http://ecos.sourceware.org/getstart.html.

I executed:

       wget --passive-ftp 
ftp://ecos.sourceware.org/pub/ecos/ecos-install.tcl
              sh ecos-install.tcl

Then I compiled sprintf1.c with this new environment and the result is 
the same. The program crashes with the first line with %f format.

Eduardo

Andrew Lunn escribió:
> On Wed, Apr 09, 2008 at 03:17:42PM -0300, Eduardo Sabaj wrote:
>   
>> Andrew, the program fails with the printing of any floating point  
>> number. Anyway, what I'm printing is :
>>
>>        printf ("main: %f\n", 2.345);
>>
>> The lines below were produced by the test program in:
>>
>> 	packages/language/c/libc/stdio/current/tests/sprintf1.c
>>
>> When the program reaches the first line with a %f format it hangs:
>>
>> 	// Check 18
>> 	ret = sprintf(x, "%f", 2.5);
>>     
>
> What toolchain are you using? Is it the eCos one from
> http://ecos.sourceware.org/getstart.html
>
>      Andrew
>
>   


-- 

Ing. Eduardo Salvador Sabaj
Analista de Desarrollo

Compañía Hasar | Grupo Hasar*
*Marcos Sastre y José Ingenieros
El Talar. Pacheco
[B1618CSD] Buenos Aires. Argentina
Tel [54 11] 4117 8900 | Fax [54 11] 4117 8998
E-mail: esabaj@hasar.com
Visítenos en: www.hasar.com
<http://www.hasar.com>Información legal y política de confidencialidad: 
www.grupohasar.com/disclaimer <http://www.grupohasar.com/disclaimer>


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] [Fwd: [ECOS] printf and floating point emulation]

[Andrew Lunn]

On Thu, Apr 10, 2008 at 10:57:32AM -0300, Eduardo Sabaj wrote:
> I'm working with one of the pre-built toolchains for Intel x386 which  
> I've downloaded some years ago. Anyway, yesterday I downloaded this same  
> toolchain from  http://ecos.sourceware.org/getstart.html.
>
> I executed:
>
>       wget --passive-ftp  
> ftp://ecos.sourceware.org/pub/ecos/ecos-install.tcl
>              sh ecos-install.tcl
>
> Then I compiled sprintf1.c with this new environment and the result is  
> the same. The program crashes with the first line with %f format.

Humm. The only ideas i have are

1) Disassemble the code and see if it is still using the FPU.

2) Make sure your libgcc1.a was build with the same flags as eCos. This
   library from the tool chain must also not use the FPU. 

   Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss