On Fri, Aug 18, 2006 at 01:17:22PM +0300, Ville Skyttä wrote:
On Thu, 2006-08-17 at 13:40 +0200, Axel Thimm wrote:
> usage of $1 in kmods'scriplets is broken.
Why would it be necessary to use it in the first place?
Don't you see any neccessity in having kmdls able to distinguish
between being first time installed/upgraded/removed?
I'm not making use of $1 in kmdls, but I can very well think of some
use cases:
o storage drivers that need to place themselves in initrd (actually
that is a missing feature often reported on qla modules at ATrpms)
o creating device drivers for systems w/o dynamic /dev (like RHEL3)
o signaling a daemon to either start/reload firmware/stop like the
case with ipw3945 & daemon - additionally needs a check whether the
operation is for the running kernel. This is also an unfinshed item
of ATrpms' kmdls for ipw3945 (there is additionally a racing during
installation of kmdls and the daemon and needs to be addressed
upstream, but that's off-topic)
You can think of other situations where you do want to know whether
this package is upgrade/first-time installed/removed. But to be fair
exactly because $1 is not in present use I rated this bug low in the
ratings column.
Dropping mechanisms to support a broken scheme just because you
don't yet have a current need for that mechanism is wrong.
--
Axel.Thimm at
ATrpms.net