On Mon, Oct 13 2008, Frans Pop wrote: > Manoj wrote: >> It correctly installs firmware in a versioned location under >> /lib/firmware. > > Problem is that even though this may avoid package conflicts, it is > *not* correct given the way Debian currently looks for firmware. Also, > it is not the way upstream recommends to install firmware.
I think where Debian looks for firmware could change. Or people can use hardlinks. > All (non-free) Debian firmware packages install the firmware files > directly under /lib/firmware, not /lib/firmware/$kernel-version. Which, of course, has nothing to do with kernel images prduced by make-kpkg, incidentally. > And Debian's udev does not consider the kernel-version when looking for > firmware. /lib/udev/hotplug.functions has: > FIRMWARE_DIRS='/lib/firmware /usr/local/lib/firmware > /usr/lib/hotplug/firmware' Seems like upstream udev firmware loader does look at /lib/firmware/$(uname -r)/, which seems sane. Why was this removed? > This means that if you start installing the same firmware file under > versioned directories, udev will use the first one it finds. Which > will be the one for some $random kernel version and not the one for > the currently running kernel. This is not a sound argument. And if I have multiple kernels installed, and only one firmware, theb the firmware on the machine will be the random firmware most recently installed. By just reverting back to upstream behaviour, this randomness in the face of versioned dirs disappears. > Installing firmware in versioned directories will also result in > bloated initramfs initrds as the whole /lib/firmware dir will be > copied into them, including the duplicated firmware files for all > installed kernel versions. Again, seems like an easy fix for the initram creator. Load files from lib/firmware. and lib/firmware/$(uname -r) > The correct solution for this would IMO be to have kernel-package > build a *separate* package for firmware which does not include the > kernel version in its package name. A package for a single firmware blob? > IMO this is an RC bug in the new kernel-package. That just makes me question your technical judgement. manoj -- The more you complain, the longer God lets you live. Manoj Srivastava <[EMAIL PROTECTED]> <http://www.debian.org/~srivasta/> 1024D/BF24424C print 4966 F272 D093 B493 410B 924B 21BA DABB BF24 424C -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]