sscanf exposed bugs

Dave Korn
Thu Apr 28 16:57:00 GMT 2005

----Original Message----
>From: Steven J Abner
>Sent: 28 April 2005 14:41

> Sorry for the intrusion, I don't have newlib installed, but I do follow
> the code.
> I happen to be working on a different version of the scanf() family.
> The test that was
> just used for the patch was great! However, upon closer inspection of
> the scanf code,
> would there possibly be missing code for "infinity" ???  I could
> convert and test your
> code but thought it might be faster if you tested for the infinity
> cases yourself.
>     Again, sorry for the intrusion.
> Steve 

  Hi Steve, I assume your post is vaguely directed in my direction?

  Well, I think you're probably right.  I've only been fixing bugs in this
code, I didn't write it in the first place and so I'm not *very* familiar
with it, but after taking a close look at the spec it seems that the newlib
sscanf _will_ fail to recognize "inf" or "infinity", and that it *is*
supposed to if it wants to be fully compliant.  As a matter of fact, it's
also legitimate to append (pretty much) any alphanumeric chars you like to
the text "nan" when representing a NaN, and sscanf is supposed to accept
them.  So I think that the sscanf implementation may still be missing a
feature or two.

Can't think of a witty .sigline today....

More information about the Newlib mailing list