Tracker to follow all ebuild conversions to linux-info / linux-mod Currently, portage provides the ${KV} variable for the kernel version. This is going away. Any ebuilds relying on this should inherit linux-info and use ${KV_FULL}. Any kernel module (or other) ebuilds using kernel-mod or kmod should be migrated to linux-mod. For sample ebuilds see nvidia-kernel and alsa-driver.
for consideration: add a function to provide for adding device nodes regardless of devfs, udev or no device management methods.
for consideration: add a function similar to kernel_is that test kernel version as follows: kernel_is_ge() { ... if [ -n "$1" ]; then [ ${KV_MAJOR} -lt $1 ] && return 1 [ ${KV_MAJOR} -gt $1 ] && return 0 RESULT=1 fi ... return ${RESULT} }
as of my next commit, kernel_is will allow ge/gt/le/lt arguments. for example: kernel_is ge 2 6 6 will be true if kernel in KERNEL_DIR is greater than, or equal to 2.6.6 if kernel_is 2 6 6 gets passed, it defaults to -eq so unless the kernel running is 2.6.6 it fails. Obviously, kernel_is 2 6, still works as normal.
gt and lt operators does not perform as expected. for instance, if kernel version is 2.6.9, kernel_is gt 2.6.8 will return false. my 2 cents: these operators should not be implemented. what would you return in case of kernel_is gt 2 6? either you force all version parts to be passed to kernel_is function or don't implement gt/lt at all.
the gt/lt aren't great. they're only going to work on KV_MAJOR, or the smallest delimeter without me changing the logic some. It isnt really that useful right now, but bare with me till after this evening. I comitted as is for ge/le really.
ok, fixed in cvs. gt/lt should now work as expected.
ummm.. no, it is not ok. try to compare 2.11.2 with 2.1.12 ;) you cannot resume to a single comparision unless you decide that every parts of version number will never have more than n digits and you precede these numbers with zeros (excepting KV_MAJOR of course - or else bash will believe it is an octal number). Your previous solution (comment #3) was the good one, but without lt/gt operators. They are useless anyway.
I've corrected the function. Hope I didn't pissed you off. You're not arround at this time and that function was broken.
btw I'm doing bestcrypt now. Thoughts on comment #1?
bestcrypt-1.5_p9 is linux-mod now. If I've done anything bad let me know.
fyi: nearly all dependancies have been fixed, except bug #74420
Here's a list of ebuilds using kernel-mod.eclass: sci-misc/comedi-0.7.68 sci-misc/comedi-0.7.67 net-firewall/tuxfrw-2.58-r1 media-libs/svgalib-1.9.19-r3 sys-apps/realtime-lsm-0.8.2_pre20041022 sys-apps/realtime-lsm-0.8.3 sys-apps/realtime-lsm-0.8.5 net-misc/cisco-vpnclient-3des-4.6.02.0030 net-misc/cisco-vpnclient-3des-4.0.5 net-misc/cisco-vpnclient-3des-4.6.00.0045-r1 net-misc/cisco-vpnclient-3des-4.0.3b-r4 net-misc/cisco-vpnclient-3des-4.0.1a-r1 net-misc/zaptel-1.0.3 net-misc/zaptel-1.0.0 net-misc/zaptel-1.0.2 net-misc/zaptel-1.0.1 app-laptop/acerhk-0.5.18 net-wireless/hostap-driver-0.2.5-r1 sys-fs/cryptsetup-0.1 media-sound/emu10k1-0.20a-r7 media-sound/emu10k1-0.20a-r6 media-sound/emu10k1-0.20a-r5 media-sound/alsa-driver-1.0.3 media-video/mplayer-1.0_pre7 media-video/mplayer-1.0_pre6-r4 media-video/mplayer-1.0_pre6-r6 media-video/mplayer-1.0_pre6-r5 media-tv/linuxtv-dvb-1.1.1-r1 media-tv/rivatv-0.8.5-r2
I'm going to convert the following ebuilds to linux-mod: media-tv/rivatv net-misc/zaptel sci-misc/comedi sys-apps/realtime-lsm
Please convert your ebuilds to linux-mod/-info: You can get help in the #gentoo-kernel IRC channel sci-misc/comedi/comedi-0.7.68.ebuild: herd: no-herd dev: caleb media-video/mplayer/mplayer-1.0_pre7.ebuild: herd: video net-firewall/tuxfrw/tuxfrw-2.58-r1.ebuild: herd: netmon dev: angusyoung
net-firewall/tuxfrw should be ok now.
tuxfrw-2.58-r1.ebuild? Doesn't look like it to me, checks should be done in pkg_setup, and you should use the linux-info functions for checking kernel version
Daniel, I just committed tuxfrw-2.61 today and I'm already requesting 2.58-r1 to be removed from other archs. It should be better now.
mplayer does not use kernel-mod/linux-mod in the latest version The only ebuild missing is: sci-misc/comedi/comedi-0.7.68.ebuild: herd: no-herd dev: caleb caleb or Leandro, please fix it.
I have removed the versions of the unmaintained sys-apps/realtime-lsm ebuild which used the deprecated kernel-mod.eclass.
qc-usb uses linux-mod but not in a sane way. We need to fix that at some point.
qc-usb is fine now, thanks to whoever fixed it
finally this is done :) Thanks for your patience
