Re: [avr-libc-dev] Bug in math library

[Colin O'Flynn]

Hello,

I have attached two patches to this message - one to patch the division 
routine, and one to patch the merge routine (needs a patch to deal with NaN 
properly).

I would appreciate any testing people can do - make sure when you build the 
avr-libc again you get all the libm.a replaced, as for me for some reason the 
libm.a in the avr architecture sub-directories didn't get replaced (ie: 
avr5\libm.a) which caused some grief...

This isn't the end of it yet - these two patches just fix the x/0 and 0/x 
problems, I think there is more problems in the library of a similar fashion. 
As well I am not too confident of the patches yet as I don't have a good 
understanding of floating-point stuff yet. I'll run lots of tests though to 
make sure... 

Regards,

 -Colin

fp_20050108.patch1
Description: Text Data

fp_20050108.patch2
Description: Text Data