gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Sam Ravnborg]

On Mon, Oct 13, 2003 at 07:34:46PM +0200, Olaf Hering wrote:
> a longstanding bug, should probably go to the main Makefile. But I dont
> know if all supported archs know about -msoft-float.

Could you please elaborate about what this fixes.
I'm very resistant to add new flags unconditionally to gcc at this stage.

	Sam

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Olaf Hering]

 On Mon, Oct 13, Sam Ravnborg wrote:

> On Mon, Oct 13, 2003 at 07:34:46PM +0200, Olaf Hering wrote:
> > a longstanding bug, should probably go to the main Makefile. But I dont
> > know if all supported archs know about -msoft-float.
> 
> Could you please elaborate about what this fixes.
> I'm very resistant to add new flags unconditionally to gcc at this stage.

Is floating point in the kernel really allowed on i386? If so, please
please add a commet to this Makefile about this fact. 

test7bk3 results, allyesconfig:


drivers/built-in.o(.text+0x2ba129): In function `amd8111e_resume':
drivers/net/amd8111e.c:1700: undefined reference to `__floatsidf'
drivers/built-in.o(.text+0x2ba145):drivers/net/amd8111e.c:1700: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x2ba14d):drivers/net/amd8111e.c:1700: undefined reference to `__fixunsdfsi'
drivers/built-in.o(.text+0x2ba1c6):drivers/net/amd8111e.c:1702: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x2ba2f0): In function `amd8111e_config_ipg':
drivers/net/amd8111e.c:1775: undefined reference to `__floatsidf'
drivers/built-in.o(.text+0x2ba308):drivers/net/amd8111e.c:1775: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x2ba310):drivers/net/amd8111e.c:1775: undefined reference to `__fixunsdfsi'
drivers/built-in.o(.text+0x2ba343):drivers/net/amd8111e.c:1778: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x2ba737): In function `amd8111e_probe_one':
drivers/net/amd8111e.c:1911: undefined reference to `__floatsidf'
drivers/built-in.o(.text+0x2ba753):drivers/net/amd8111e.c:1911: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x2ba75b):drivers/net/amd8111e.c:1911: undefined reference to `__fixunsdfsi'
drivers/built-in.o(.text+0x2ba838):drivers/net/amd8111e.c:1940: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x407fd7): In function `tuner_set_tv_freq':
drivers/media/dvb/ttpci/av7110.c:2709: undefined reference to `__floatsidf'
drivers/built-in.o(.text+0x407ff3):drivers/media/dvb/ttpci/av7110.c:2709: undefined reference to `__ltdf2'
drivers/built-in.o(.text+0x408001):drivers/media/dvb/ttpci/av7110.c:2711: undefined reference to `__floatsidf'
drivers/built-in.o(.text+0x408019):drivers/media/dvb/ttpci/av7110.c:2711: undefined reference to `__ltdf2'
drivers/built-in.o(.text+0x408060):drivers/media/dvb/ttpci/av7110.c:2720: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x408076):drivers/media/dvb/ttpci/av7110.c:2720: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x5c24d0): In function `sisfb_do_set_var':
drivers/video/sis/sis_main.c:654: undefined reference to `__floatsidf'
drivers/built-in.o(.text+0x5c24ec):drivers/video/sis/sis_main.c:654: undefined reference to `__divdf3'
drivers/built-in.o(.text+0x5c24f7):drivers/video/sis/sis_main.c:655: undefined reference to `__floatsidf'
drivers/built-in.o(.text+0x5c2510):drivers/video/sis/sis_main.c:655: undefined reference to `__divdf3'
drivers/built-in.o(.text+0x5c251b):drivers/video/sis/sis_main.c:656: undefined reference to `__floatsidf'
drivers/built-in.o(.text+0x5c2534):drivers/video/sis/sis_main.c:656: undefined reference to `__divdf3'
drivers/built-in.o(.text+0x5c2540):drivers/video/sis/sis_main.c:656: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x5c2554):drivers/video/sis/sis_main.c:656: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x5c255c):drivers/video/sis/sis_main.c:656: undefined reference to `__fixunsdfsi'
drivers/built-in.o(.text+0x5c28b8):drivers/video/sis/sis_main.c:675: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x5c28d1):drivers/video/sis/sis_main.c:675: undefined reference to `__adddf3'
drivers/built-in.o(.text+0x5c28ea):drivers/video/sis/sis_main.c:675: undefined reference to `__adddf3'
drivers/built-in.o(.init.text+0x6252d): In function `sisfb_init':
drivers/video/sis/sis_main.c:4450: undefined reference to `__floatsidf'
drivers/built-in.o(.init.text+0x6253f):drivers/video/sis/sis_main.c:4450: undefined reference to `__divdf3'
drivers/built-in.o(.init.text+0x62547):drivers/video/sis/sis_main.c:4450: undefined reference to `__fixunsdfsi'


thanks.

-- 
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Thomas Winischhofer]

You wrote:
 >drivers/built-in.o(.text+0x5c24f7):drivers/video/sis/sis_main.c:655: 
 >undefined \
 >reference to `__floatsidf' \
 >drivers/built-in.o(.text+0x5c2510):drivers/video/sis/sis_main.c:655: 
 >undefined \
 >reference to `__divdf3' \
 >drivers/built-in.o(.text+0x5c251b):drivers/video/sis/sis_main.c:656: 
 >undefined \
 >reference to `__floatsidf' \
 >drivers/built-in.o(.text+0x5c2534):drivers/video/sis/sis_main.c:656: 
 >undefined \
 >reference to `__divdf3' \
 >drivers/built-in.o(.text+0x5c2540):drivers/video/sis/sis_main.c:656: 
 >undefined \
 >reference to `__adddf3' \
 >drivers/built-in.o(.text+0x5c2554):drivers/video/sis/sis_main.c:656: 
 >undefined \
 >reference to `__adddf3' \
 >drivers/built-in.o(.text+0x5c255c):drivers/video/sis/sis_main.c:656: 
 >undefined \
 >reference to `__fixunsdfsi' \
 >drivers/built-in.o(.text+0x5c28b8):drivers/video/sis/sis_main.c:675: 
 >undefined \
 >reference to `__adddf3' \
 >drivers/built-in.o(.text+0x5c28d1):drivers/video/sis/sis_main.c:675: 
 >undefined \
 >reference to `__adddf3' \
 >drivers/built-in.o(.text+0x5c28ea):drivers/video/sis/sis_main.c:675: 
 >undefined \
 >reference to `__adddf3' drivers/built-in.o(.init.text+0x6252d): In 
 >function \
 >`sisfb_init': drivers/video/sis/sis_main.c:4450: undefined reference 
to >`__floatsidf'
 >drivers/built-in.o(.init.text+0x6253f):drivers/video/sis/sis_main.c:4450: >undefined \
 >reference to `__divdf3' \
 >drivers/built-in.o(.init.text+0x62547):drivers/video/sis/sis_main.c:4450: >undefined \
 >reference to `__fixunsdfsi'


This is fixed in the latest version of sisfb I sent to James Simmons; 
one more reason to merge the fbdev stuff...

Thomas

-- 
Thomas Winischhofer
Vienna/Austria
thomas AT winischhofer DOT net          *** http://www.winischhofer.net/
twini AT xfree86 DOT org

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 1745 bytes --]

On Tue, 2003-10-14 at 10:12, Olaf Hering wrote:
>  On Mon, Oct 13, Sam Ravnborg wrote:
> 
> > On Mon, Oct 13, 2003 at 07:34:46PM +0200, Olaf Hering wrote:
> > > a longstanding bug, should probably go to the main Makefile. But I dont
> > > know if all supported archs know about -msoft-float.
> > 
> > Could you please elaborate about what this fixes.
> > I'm very resistant to add new flags unconditionally to gcc at this stage.
> 
> Is floating point in the kernel really allowed on i386? If so, please
> please add a commet to this Makefile about this fact. 
> 
> test7bk3 results, allyesconfig:
> 
> 
> drivers/built-in.o(.text+0x2ba129): In function `amd8111e_resume':
> drivers/net/amd8111e.c:1700: undefined reference to `__floatsidf'

real bug:
        if(lp->options & OPTION_DYN_IPG_ENABLE)
                mod_timer(&lp->ipg_data.ipg_timer,
                                jiffies + (IPG_CONVERGE_TIME * HZ));
 
where 
#define      IPG_CONVERGE_TIME 0.5

> drivers/media/dvb/ttpci/av7110.c:2709: undefined reference to `__floatsidf'

worse:
       if (freq < 16*168.25 )
                config = 0xa0;
        else if (freq < 16*447.25)
                config = 0x90;
        else
                config = 0x30;

> drivers/built-in.o(.text+0x5c24d0): In function `sisfb_do_set_var':
> drivers/video/sis/sis_main.c:654: undefined reference to `__floatsidf'


static int sisfb_do_set_var(struct fb_var_screeninfo *var, int isactive,
                      struct fb_info *info)
{
        unsigned int htotal =
                var->left_margin + var->xres + var->right_margin +
                var->hsync_len;
        unsigned int vtotal = 0;
        double drate = 0, hrate = 0;
 
ugh


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Olaf Hering]

 On Tue, Oct 14, Arjan van de Ven wrote:

> ugh

My argument is that this shouldnt be done the usual opensource way
'someone else will fix my shit'. Instead, the driver authors should
notice their mistakes right away.

Linus, please add -msoft-float to test8 and keep it. Thanks.


-- 
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Thomas Winischhofer]

Olaf Hering wrote:
 > On Tue, Oct 14, Arjan van de Ven wrote:
 >
 > > ugh
 >
 > My argument is that this shouldnt be done the usual opensource way
 > 'someone else will fix my shit'. Instead, the driver authors should
 > notice their mistakes right away.

I will certainly not maintain the ancient driver in current 2.6 vanilla. 
This is a framebuffer driver and like all fbdev-related stuff it is 
properly maintained in the fbdev-tree, waiting to merged into mainline.

Thomas

-- 
Thomas Winischhofer
Vienna/Austria
thomas AT winischhofer DOT net          *** http://www.winischhofer.net/
twini AT xfree86 DOT org

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Andrew Morton]

Thomas Winischhofer <thomas@winischhofer.net> wrote:
>
> This is fixed in the latest version of sisfb I sent to James Simmons; 
>  one more reason to merge the fbdev stuff...

Well afaik nobody is sending anything.  I've had some 35,000 line fbdev
patch in -mm for six weeks but have not seen anything since.

What's up?

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Thomas Winischhofer]

Andrew Morton wrote:
> Thomas Winischhofer <thomas@winischhofer.net> wrote:
> 
>>This is fixed in the latest version of sisfb I sent to James Simmons; 
>> one more reason to merge the fbdev stuff...
> 
> 
> Well afaik nobody is sending anything.  I've had some 35,000 line fbdev
> patch in -mm for six weeks but have not seen anything since.
> 
> What's up?

I think the only one to answer this is James...

Thomas

-- 
Thomas Winischhofer
Vienna/Austria
thomas AT winischhofer DOT net          *** http://www.winischhofer.net/
twini AT xfree86 DOT org

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Meelis Roos]

TW> This is a framebuffer driver and like all fbdev-related stuff it is 
TW> properly maintained in the fbdev-tree, waiting to merged into mainline.

Since James seems busy, someone should step up, split these changes into
patches, test (or let people test) them separately and submit to kernel.
Of course in coordination with James, he knows hat should be stable and
what not. I would take this myself by have not enough time.

-- 
Meelis Roos (mroos@linux.ee)

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Thomas Winischhofer]

Meelis Roos wrote:
> TW> This is a framebuffer driver and like all fbdev-related stuff it is 
> TW> properly maintained in the fbdev-tree, waiting to merged into mainline.
> 
> Since James seems busy, someone should step up, split these changes into
> patches, test (or let people test) them separately and submit to kernel.
> Of course in coordination with James, he knows hat should be stable and
> what not. I would take this myself by have not enough time.

That sounds a little like "Linus is busy, let someone else quickly hop 
in"... :)

The fbdev tree is quite well tested AFAIK, and I am sure James will 
respond soon. He can't be too busy, as he was working on the tree yesterday.

Thomas


-- 
Thomas Winischhofer
Vienna/Austria
thomas AT winischhofer DOT net          *** http://www.winischhofer.net/
twini AT xfree86 DOT org

Re: gcc -msoft-float [Was: Linux 2.6.0-test7 - stability freeze]

[Geert Uytterhoeven]

On Tue, 14 Oct 2003, Thomas Winischhofer wrote:
> Meelis Roos wrote:
> > TW> This is a framebuffer driver and like all fbdev-related stuff it is 
> > TW> properly maintained in the fbdev-tree, waiting to merged into mainline.
> > 
> > Since James seems busy, someone should step up, split these changes into
> > patches, test (or let people test) them separately and submit to kernel.
> > Of course in coordination with James, he knows hat should be stable and
> > what not. I would take this myself by have not enough time.
> 
> That sounds a little like "Linus is busy, let someone else quickly hop 
> in"... :)
> 
> The fbdev tree is quite well tested AFAIK, and I am sure James will 
> respond soon. He can't be too busy, as he was working on the tree yesterday.

Indeed. James said he was going to submit the patches after he has fixed a
critical problem with the soft cursor code.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds