Re: [rfc/rfa] Almost eliminate HOST_{FLOAT,DOUBLE,...}_FORMAT

Re: [rfc/rfa] Almost eliminate HOST_{FLOAT,DOUBLE,...}_FORMAT

[Kevin Buettner]

On Jun 29,  1:23pm, Andrew Cagney wrote:

> This patch changes GDB so that, if the host configury didn't specify the 
> floating point format, it will fall back immediatly to floatformat_*(). 
>   Previously, GDB would assume that the HOST had IEEE FP and try to use 
> that.
> 
> For hosts that don't specify a HOST_*_FORMAT, this will mean a loss of 
> FP precision when using GDB :-/

Are you suggesting that hosts should define HOST_*_FORMAT?  Or should
we attempt to eliminate the HOST_*_FORMAT macros entirely?

(The only HOST_*_FORMAT macro currently defined is HOST_LONG_DOUBLE_FORMAT
and it is only defined by some i386 hosts.)

BTW, you'll need to address the use of HOST_FLOAT_FORMAT in
arm_linux_push_arguments() in arm-linux-tdep.c or else that file
won't compile any longer after eliminating the defines from defs.h.

Kevin

Re: [rfc/rfa] Almost eliminate HOST_{FLOAT,DOUBLE,...}_FORMAT

[Kevin Buettner]

On Jun 29,  1:48pm, Kevin Buettner wrote:

> BTW, you'll need to address the use of HOST_FLOAT_FORMAT in
> arm_linux_push_arguments() in arm-linux-tdep.c or else that file
> won't compile any longer after eliminating the defines from defs.h.

Oops.  Never mind.  I'm behind on email and just got to some of the
earlier mail where you fixed this problem on ARM.

Kevin

Re: [rfc/rfa] Almost eliminate HOST_{FLOAT,DOUBLE,...}_FORMAT

[Andrew Cagney]

> On Jun 29,  1:23pm, Andrew Cagney wrote:
> 
> 
>> This patch changes GDB so that, if the host configury didn't specify the 
>> floating point format, it will fall back immediatly to floatformat_*(). 
>> Previously, GDB would assume that the HOST had IEEE FP and try to use 
>> that.
>> 
>> For hosts that don't specify a HOST_*_FORMAT, this will mean a loss of 
>> FP precision when using GDB :-/
> 
> 
> Are you suggesting that hosts should define HOST_*_FORMAT?  Or should
> we attempt to eliminate the HOST_*_FORMAT macros entirely?


I think we should eliminate HOST_*_FORMAT entirely.  GDB should always 
use something like ``sim/common/sim-fpu.[hc]'' to do its FP emulation. 
that way, no matter what the host/target, GDB provides both consistent 
and exact FP behavour.


	Andrew