From: Junio C Hamano <junkio@cox.net>
Date: Thu, 15 Sep 2005 08:17:16 GMT
Subject: Re: git-diff-tree rename detection bug
Message-ID: <7vll1y9243.fsf@assigned-by-dhcp.cox.net>
In-Reply-To: <17192.65054.520959.454610@cargo.ozlabs.ibm.com>

Paul Mackerras <paulus@samba.org> writes:

> The strlen() in glibc for ppc is unbearably clever hand-coded
> assembly, which loads up 8 bytes at a time (once it has the address
> 8-byte aligned), and does various ANDs and ORs and ADDs and
> conditional branches.  If some of the 8 bytes aren't defined, it will
> in many cases branch one way or the other based on the undefined
> bytes, but end up computing the same result on either branch.

This reminds me of what I did in my previous life, writing a
memory allocation checker -- this was before Valgrind -- and
found out that strcpy in the C library that came with Solaris
had a similar clever trick.  What was interesting was that
copying a string starting at the (PAGESIZE-3)th byte on a page
and NUL terminated at the end of the same page ended up
prefetching the first word from the next page (please do not ask
me about the details -- I do not remember the disassembly of
that part of the code anymore).  It was not an inconvenience for
our memory checker but was a real bug -- the next page could
very well be unaccessible.

The bug was fixed in the next version of the C library when we
updated our Solaris box.