[Clfs-dev] Recent Updates to CLFS builds

[William Harrington]

Some more updates since I've gone through the pure64 bit book.

I've been trying different udev configurations that make sense, but  
our udev bootscript isn't matching.

Our bootscript expects /lib/udev/udevadm   (that is fixed with -- 
libexecdir=/lib)
                           expects /bin/udevadm              (not a  
problem for now)

We do need to update our bootscript for udev; however. I think  
udevadm needs to be in /sbin and so the udev bootscript would need to  
be updated since it runs: /bin/udevadm and not /sbin/udevadm like it  
should be.

What is the community's thought on this?

I used this configure option the last round:

--prefix=/usr --with-rootprefix="" --sysconfdir=/etc --libexecdir=/ 
lib --enable-rule_generator
--disable-introspection --disable-keymap --disable-gudev --with-usb- 
ids-path=no
--with-pci-ids-path=no

Althouh this will put libudev into /usr/lib.   Do we need to make  
sure the libs get put into /lib in case someone has /usr as a  
separate partition and it is not yet mounted when the udev script is  
ran?

I was looking to see what all may be linked to /usr/lib/libudev  
libraries but I don't see anything that would harm that. But I would  
like to make all of this proper.


============================NCURSES===============

When I've been configuring ncurses, I made sure the terminfo install  
directory was correct.

/tools/share/terminfo
/usr/share/terminfo

however, with the latest ncurses branch update patch, there is  
apparently another configure option that needs to be made as  
terminfos aren't found. They are being looked in /share rather than / 
usr/share.  So we need to investigate and adjust the book accordingly.

=================BOOTING================

DEVTMPFS issues.

I've had issues with this and the bootscripts as well.

I always have to have devtmpfs options in the kernel to automount.
I do this because the filesystem is mounted readonly and the udev script
cannot copy the static files as it is still read only. It's supposed  
to be read only
with tmpfs mounted at /dev so static nodes can be copied. Well the  
boot will fail cause of this.

Is our bootscript of mounting virtual filesystems correct in the  
order of things?

If I have devtmpfs automount, it is not an issue as /dev is populated  
with devices and then can
check filesystems as the block device nodes are there.

If not, the boot fails becuase filesystem cannot be checked as there  
are no block device nodes available.

with journaling filesystems I set to not check anyway, but still...  
it's a possible problem with the order our bootscript does things. I  
can't pinpoint it.

That is it for now.

Sincerely,

William Harrington