fusermount3 get's skipped if not setuid, which breaks new fuse implementations

[thkukuk]

PR#82 skips fusermount3 if the setuid bit is not set.

Some Linux distributions working currently on make it possible to run a Linux distribution with "NoNewPrivs" enabled. For this, fuse3 get's replaced with defused, a fusermount3 replacement working without setuid bit.
AppImage does not work with this, since the fusermount3 replacement does not have a setuid bit set. Contrary, make it setuid would be a security problem, since that would break the authentication stuff.

Can we remove that check for the setuid bit?

[probonopd]

Thanks for bringing this to our attention.

Details

Aargh. why is eveything always being fiddled with and not left alone indefinitely once it is feature complete in Linux land....

I only wish we had a stable system that we can rely on over decades. But no. First libfuse2 -> libfuse3 with 0 months transition time in which both would have been installed in Ubuntu, and now this.

[probonopd]

Had a look at defused and am not thrilled by what I saw. It seems to assume many things like libsysemd, D-Bus, polkit, JavaScript... which is not desirable outside of the Fedora/glib ecosystem. I definitely would not want to need these things on my system just to be able to unmount FUSE mounts. FreeBSD does this so much better.

I have opened an upstream issue:

• Drop libsystemd, D-Bus and polkit dependencies Skyb0rg007/defused#67

@thkukuk if all that would be needed for it to work is loosening the setuid check, I'd not be against it but in the form:
"Check that it is either setuid and/or defused". I don't have a way to test it though, since I won't be letting libsysemd, D-Bus, polkit onto my system.

[Skyb0rg007]

Loosening the setuid check would be great!

One option for maximal backwards compatibility is to perform the search for a setuid binary as normal, but additionally keep a reference to the first binary which is executable. After the search finishes, if no setuid binaries were found, then the binary which is just executable will be returned.

The environment variable FUSERMOUNT_PROG bypasses the setuid checks though, so maybe I can just recommend users set that.

[probonopd]

Please test whether the PR fixes it. Thanks!