mimosa wrote:To put it another way, these ergonomically impressive tools (like Lilosetup or the installer) already go out of their way to make things easier than they need be. I'm just playing with ideas for ways to take that further - up to the devs whether, on balance, that's workable and desirable. My feeling is still that a sanity check on /etc/fstab and a warning if appropriate would be worth considering.
A sanity check can't be reliable in any case. As akuna pointed out this is out of the scope of the lilosetup tool.
How should I check whether the fstab is valid? It's just not possible unless you're running in the system, which the fstab is written for. What to check anyway? Device existence? User mistakes swapping device names? The latter is impossible the first can't be in any way reliable on a different system. Devices may exist or may not due to missing kernel/software support. They might even change their order!!! So only UUIDs could allow more reliable (by far not reliable yet) checking of devices despite changing order. But as you noticed yourself, most of the problem doesn't exist with UUIDs. Hard drives can just not be plugged out and intended to fail and on top they might not have automount enabled or use some other fancy feature, which would require a parser that can't work for every version of mount. Did you ever consider raid, luks, device-mapper, lvm, zfs and other fancy fs/disk functionality?
You also mentioned that user setups tend to become increasingly complex. This only makes the situation worth, because there's even more to care about. On a sidenote imho automation is not very high with complex setups on Salix and Slackware anyway, which is not that bad at all. There's so much too care about in these setups, that it's likely to be error-prone. Not considering the ressources to implement such functionality.
In conclusion this would be a warning that's so likely to fail, that it barely annoys many users. So not soft and seamless like a kiss.