Performance testing w BackupExec, comments requested.

[Steve Langasek]

On Wed, Jan 25, 2006 at 04:14:11PM -0500, David Collier-Brown wrote:
> Jeremy wrote:
> >>>The uname(2) system call conforms to SVr4, SVID, POSIX, and X/OPEN but 
> >>>is not
> >>>available in BSD4.3.  If this call exists (configure test) and the 
> >>>"sysname"
> >>>field is "Linux" and the "release" field, parsed from a string like
> >>>2.6.8-2-686, shows the release to be greater than 2.6.15, then enable 
> >>>else
> >>>disable.

> Steve Langasek wrote:
> >>Ugh. That's *horrible* :-). Tempting though.... :-).
> >>No, this really is something vendors should address (IMHO).

>   Sun tries (;-)) We like to make the linker do this checking,
> actually. Interfaces have hidden numbers indicating what standard
> they're from, like SISCD_2.3 and SYSVABI_1.3 or SUNW_1.1
> If sendfile wasn't in the standard libs, you'd be linked
> to a software emulation, thusly:

Given that the bugs are all in the kernel implementation of sendfile, it
doesn't seem very practical to have a check at the ELF level.  I suppose you
could have glibc itself check the uname and bail when asked for sendfile on
an older kernel, but I don't imagine that the glibc folks would be keen to
implement such a check either. :)

-- 
Steve Langasek                   Give me a lever long enough and a Free OS
Debian Developer                   to set it on, and I can move the world.
vorlon at debian.org                                   http://www.debian.org/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://lists.samba.org/archive/samba-technical/attachments/20060125/022f148c/attachment.bin