[OpenAFS] 1.4.x fileserver crashes on amd64_linux2X when tracing turned on

Jeffrey Hutzelman jhutz@cmu.edu
Wed, 15 Mar 2006 12:14:26 -0500

On Tuesday, March 14, 2006 10:26:43 AM -0500 Jim Rees <rees@umich.edu> 

> I'm opposed to nested includes because it makes it difficult to diagnose
> include ordering violations but I realize that's a losing battle.

Nested includes are fine for cases where there are actual dependencies.
And there are arguments to be made both ways for software that uses the 
model where every source file includes some common header that's expected 
to include all the required system headers.  Outside of the cache manager, 
AFS mostly does not use that model.  It seems to me that it is 
inappropriate to add a new #include to afsutil.h which does not satisfy a 
dependency of that header, rather than modifying the affected source files.

>   I actually wonder why hardly anybody uses the '%I' format instead
>   which util/snprintf.c understands - perhaps because a move to IPv6 (if
>   ever) would be more complicated?

No, it wouldn't.  Both %I and inet_ntoa render ASCII representations of 
IPv4 addresses expressed as network-order integers.

> I suspect snprintf was added after most of these callers were written to
> use inet_ntoa.  They could safely be changed now but no one has bothered.

Yup.  That snprintf, and its support of %I, are both relatively new to 

-- Jeff