Bugzilla Bug 139337

Attached are several patches for catalyst to get in some final mips tweaks and
miscellaneous catalyst tweaks.  Some are safe to go right in, others could use
a review and consideration.  I'll denote which should need review.

Created an attachment (id=90980) [edit]
mips tweaks for targets/support/create-iso.sh

Changes the supported fstype to squashfs, adds some additional comments to
better explain things, and some conditional checks before shifting files
around.

Created an attachment (id=90981) [edit]
Update the source file for generating the arcload config

Modifies the arcload config source file so that the generated config passes
needed variables for loading an X-based LiveCD.

Created an attachment (id=90988) [edit]
Update gdm.conf handling to account for gdm-2.14 differences

gdm-2.14 uses defaults in /usr/share/gdm/defaults.conf and desired changes go
into an overlay config in /etc/X11/gdm/custom.conf versus the old behavior of
applying sed maginc to /etc/X11/gdm/gdm.conf.

Created an attachment (id=90989) [edit]
Prevent catalyst from re-merging sources all the time

This one needs some review.  In my test runs for both netboot2 and
livecd-stage2, catalyst would always wipe out and re-merge the kernel sources
even when kerncache was set, which got annoying.  The deleted block of code
doesn't make a whole lot of sense to me, but removing it makes catalyst keep
the sourcetree around unless -p is passed or the kerncache is manually deleted.
 Some review on this behavior might be warranted.

Created an attachment (id=90990) [edit]
Favour settings in the specfile versus make.conf

If a particular setting is specified in the spec files, use that instead fo
what may be specified in the make.conf in the chroot.  If no spec file setting
is available, then the value from make.conf is used if it exists.

This one might need some review too.

Created an attachment (id=90992) [edit]
Add sanity checks to mips' bootloader-setup section

Before moving or deleting directories, check that the target exists first.

Created an attachment (id=91008) [edit]
Fix chost/c[xx]flags/ldflags overrides

This modifies generic_stage_target so that the override functions for chost,
cflags, cxxflags, and ldflags only pull their values from the make.conf file by
default.  As such, stage3, stage4, grp, tinderbox, netboot[1|2], and the
embedded targets all use this method to set these values.  stage1 and stage2
override this behavior, however, and pull their values from the spec file.

OK.  I put these in CVS.  I did do a little shuffling of the gdm stuff because
I didn't like how it read and preferred to try to keep things to 80 columns. 
It doesn't change the functionality, except in places where the patch made
incorrect assumptions, such as defaulting to the "gentoo" user.  This will show
up in 2.0_rc50.

The gdm patch was a bit wrong. It tries to interpolate ${sedxtra} in a
single-quoted string, which just doesn't work. I modifed it a bit locally, and
it works fine.

First, I changed sedxtra to autologin_user:

                [ -n "${clst_livecd_users}" ] && [ -n "${first_user}" ] \
                    && autologin_user="${first_user}" \
                    || autologin_user="gentoo"

Then I changed the sed a bit:

                sed -e
's:\(\[daemon\]\)$:\1\nTimedLoginEnable=true\nTimedLoginDelay=10\nTimedLogin='${autologin_user}':'

Umm... Andrew?  Did you read the patch I actually *applied* versus what was
posted here?  :P

Nope :P

Fixed in 2.0_rc50...