On ltmodem-8.31_alpha10.ebuild: 106 einfo "Use /dev/tts/LT0 to access modem" 107 einfo "If you have problems, read this doc:" 108 einfo "/usr/share/doc/${PF}/html/post-install.html" Should be, at lest, on an ~x86 2.6 udev system: /dev/ttyLTM0 and no (/dev/ttyLT0) In both cases (devfs and udev), should be better point to /dev/modem. devfs rules (/etc/devfs.d/ltmodem) already creates the symlink, an udev rule should be installed too in order to put in the einfo "/dev/modem" Something like this inside /etc/udev/rules.d/ KERNEL="ttyLTM0", NAME="%k", SYMLINK="modem" or inside the "which dev file system" condition (if [ -e ${ROOT}/dev/.devfsd ]...) show the correct path to the node. /dev/tts/LT0 for devfs, and /dev/ttyLTM0 for udev, if i'm correct. :-)
Inside the tarball ltmodem-2.6-alk-7.tar.bz2 on ltmodem-2.6-alk-7/docs/ltmodem.rules there are a better udev rule. KERNEL="ttyLTM[0-9]", NAME="%k", MODE="0660", GROUP="tty", SYMLINK="modem" Also create a group named "modem" would be nice...
Also in ltmodem-8.31_alpha10.ebuild 28 MODULESD_LTMODEM_ALIASES=( "char-major-62 ltserial" 29 "/dev/tts/LT0 ltserial" 30 "/dev/modem ltserial" ) the node are "hardcoded", it should point to /dev/ttyLTM0 on udev? if [ -e ${ROOT}/dev/.devfsd ]; then MODULESD_LTMODEM_ALIASES=( "char-major-62 ltserial" "/dev/tts/LT0 ltserial" "/dev/modem ltserial" ) eend $? elif [ -e ${ROOT}/dev/.udev ]; then MODULESD_LTMODEM_ALIASES=( "char-major-62 ltserial" "/dev/ttyLTM0 ltserial" "/dev/modem ltserial" ) eend $? fi In order that linux-mod.eclass can create the corresponding /etc/modules.d/ltmodem entry. Note0: I'm not sure but seems that in some devfs systems /dev/ttyLT0 is created and not /dev/tts/LT0. (README on ltmodem-2.6-alk-7.tar.bz2) Note1: My entries inside /etc/modules.d/ltmodem points to a non existen node (/dev/tts/LT0) and it's working...
there is no point in creating a new "modem" group. why would be 2 such groups (see dialout)? see version 8.31_alpha10-r1. now udev should create /dev/ttyLTM0 device. reopen this bug if not.
Udev always created /dev/ttyLTM0 device on my system, no problem about that... now thanks to the new udev rules on 8.31_alpha10-r1 ebuild it automatically creates the /dev/modem symlink. The bug report is about the einfo information (where to access the modem and the dialout group thing), also for the correct MODULESD_LTMODEM_ALIASES and /etc/modules.d/ltmodem data. I think the einfo information should say /dev/modem instead of /dev/tts/LT0 or not? and the /etc/modules.d/ltmodem should point to the correct device (which it can change according the dev/kernel implementation) % ls -la /dev/tts/LT0 ls: /dev/tts/LT0: No such file or directory Note: About the 'modem' group, my fault, I did not notice the existence of the `dialout` group. Anyway, devfs are using it? Udev, are using it now with the new 8.31_alpha10-r1 ebuild. % ls -la /dev/ttyLTM0 crw-rw---- 1 root dialout 62, 64 Mar 5 23:34 /dev/ttyLTM0 % ls -la /dev/modem lrwxrwxrwx 1 root root 7 Mar 7 17:34 /dev/modem -> ttyLTM0 But, should the ebuild inform about that? I mean, like games.eclass do? "Remember, in order to access the modem, you have to be in the 'dialout' group."
for the aliases thing: the point of setting aliases is for creating a kind of symlink (a virtual one if you like) between a name and a module. Since udev creates the /dev/ttyLTM0, there is no point in setting an alias for it. please test it with devfs. I need to know what dev files are created. FYI: if you boot with "gentoo=noudev" and have devfs installed, devfs will be used instead of udev.
I think you owe this simple test to me. After all, you reopened this bug...
I'm compiling the kernel with devfs support, I will test it when it finish :-) I reopened beacause i think einfo should inform about the correct device to use, if i'm wrong it could be marked invalid or something. :-) Sorry I did'nt had the time in the weekend, I'm back right now.. Thanks!
I've added warnings about device and /var/lock acess to postinst.
