On Sat, Sep 11, 2004 at 03:27:01PM +0200, Fabrice LORRAIN wrote: > Hello, > > On an uptodate pre-sarge box, the following shows some binaries linked > with libraries under /usr. Those binaries might (didn't test) fail when > /usr is not mounted. > > Is there a policy against this ?
We follow the FHS, and it doesn't have a strict[1][2] word about it, but as far as I can see, either these binaries are not needed before /usr is mounted, in which case they should be moved to /usr/[s]bin, or they are needed, in which case they are currently broken. With the exception of fsck.cramfs, I dont see that any of those binaries is used before S35mountall.sh@ in /etc/rcS.d, so they seem to be the former case. And, excuse me, what the hell is the point of fsck for a read-only filesystem anyway? > Should I feel bugs ? That would be mass filing[3], which you need to get consensus on debian-devel first. However, I think the preferred aproach for this would be to send patches for linda/lintian to identify any such linking in packages. This way maintainers will notice the odd linking int their packages when they run lintian/linda before the next upload they do. After automatic methods to identify /[s]bin/ binaries linking to /usr/lib have existed for a while, a mass filing process can be started. [1] http://www.pathname.com/fhs/pub/fhs-2.3.html#BINESSENTIALUSERCOMMANDBINARIES [2] http://www.pathname.com/fhs/pub/fhs-2.3.html#SBINSYSTEMBINARIES [3] http://www.debian.org/doc/developers-reference/ch-beyond-pkging.en.html#s-submit-many-bugs
