Why have a /dev/mouse link at all? Symbolic links in /dev are a good idea for like devices, so cdroms (all of which look the same to mount and freinds - just block devices) get links to /dev/cdrom. But mice don't look the same, and both X and gpm both need to be told what the mouse behind /dev/mouse is anyway, why not just modify they configuration files to tell them where to find it as well. Mouseconfig already modifies the X configuration files to set the mouse to /dev/mouse, why not just specify the exact device and be done with it? Similarly gpm does not require /dev/mouse, it can be told with -m /path/to/mouse. Fixing this will make devfs support easier (RedHat already ships devfsd and the line in rc.sysinit so users who want it can use it - at least to the stage of booting), and won't inconvenience other users. Other approches to this problem, like kudzu modifying the devfsd.conf file are silly (bug 22810), and this approch can still use the old device names - with devfsd managing the compatability links as before. The work involved is only a few lines in /etc/rc.d/init.d/gpm, /etc/sysconfig/mouse (add a line for the path to the mouse) and mouseconfig.c (put the path the the mouse, rather that /dev/mouse when rewriting the config).
Created attachment 7164 [details] patch for init.d/gpm to use a DEVICE variable and make it usable on devfs
Created attachment 7166 [details] Better version of the init.d/gpm patch
Created attachment 7167 [details] A patch to mouseconfig.c to make it read and record the device it uses, instead of using the /dev/mouse symlink
This patch implements (along with the patch to gpm's init.d entry, mirrored in bug 23500) the storage of the mouse device in /etc/sysconfig/mouse, rather than in the /dev/mouse symbolic link. This allows for the use of gpm (and presumably X) on systems with devfs. It should have no detremental effect on non-devfs users, and uses the old standard names. (devfsd provides these as a compatibility measure, and RedHat ships with devfsd - presumably for user's own sanity). My muck-about machine, on which this was developed however does not have X installed, but the changes to its configuration file seem correct. Also note that I am not yet particularly competent in C, but this patch appears to work and should at least be a good start on implmenting this feature. The use of this patch should allow RedHat machines to boot correctly into gpm-enabled console and X without issues, even if they want to try devfs.
I will have a look at this over the next few days, thanks for the patches.
Is this still being looked at?
I have just completed a text mode interface for redhat-config-mouse, which means that mouseconfig will be deprecated in the next release of Red Hat Linux. Therefore, I will not be putting any development time towards fixing mouseconfig bugs, so I'm closing this as 'wontfix'.
That's fine, but does the new redhat-config-mouse still rely on /dev/mouse? If so, can this bug be re-assigned to the new program? If not, then thankyou! Andrew Bartlett
Changed to 'CLOSED' state since 'RESOLVED' has been deprecated.