[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 13 Jul 2011 16:03:27 -0400
From: Rich Felker <dalias@...ifal.cx>
To: musl@...ts.openwall.com
Subject: Re: cluts review
On Wed, Jul 13, 2011 at 06:55:01PM +0200, Luka Marčetić wrote:
> that gets executed with each string). The huge long long integer
> printout is I guess what makes it look as if there are more messages
> than there are. Other than that, it's probably the implementations.
musl definitely has a lot of bugs in this area, and may rewrite the
code (for floating point I will definitely rewrite it because the
current approach is inexact and fundamentally can't be right). But I
think at least the huge long double test is wrong:
strtold(.... offsets endptr by 4940 instead of by 0
As far as I can tell, the string you pass is of the expected form for
the subject sequence. It may overflow to infinity or lose precision,
but it's still valid. strtold is only supposed to store the initial
pointer in endptr if the subject sequence is empty or does not have
the expected form; otherwise it stores a pointer to the byte just past
the end of the subject sequence.
Does this make sense?
Admittedly I need to review the rest of the tests more thoroughly too.
Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
