The reason not to add server cleanup right into
distro/server-base is the possibility to build
combined installers (read centaurus).
Gotta sort it out some day...

Applying set() and only set() to a GLOBAL_* is safe
but still a potentially confusing example; so let's
just do it right (and warn unsuspecting folks too).

This was done while debugging GLOBAL_CLEANUP_PACKAGES
getting doubled... as it got no hard initial value but
rather was always added to, it appeared that at least
stage2/Makefile would obtain a once-initialized value
from upstream and double it while including distcfg.mk
itself.

It's way less hassle to just proxy the value once.

This is a sort of anti-feature which removes and not builds;
still with mkimage-profiles' approach we can at least build
up the removal procedures as well.

It's what triggered the tagged scripts, BTW.

From now on a feature can contain this tree:

.
+- scripts.d/
+- image-scripts.d/
`- tagged/
   +- scripts.d/
   `- image-scripts.d/

...per subprofile part or in its root -- the latter one
gets merged into toplevel directory responsible for the
final image build.

NB: autoselected tags include only subprofile names
    (or both parts, for complex subprofiles) --
    this is highly prone to change yet!

Mostly needed for tags2lists at the moment;
note that PATH is exported for make recipes
but has no effect on make functions.

$(IMAGES) in prerequisites definitely won't work,
so let's do it straightforward.

The docs would be purged anyways, let's not install them either.

Somewhat more uniform style and a few additional bits.

Introduced support for hooks to be added to every
derivative substage of a "base" stage (think stage2/*).

Specific hooks for e.g. stage2/live would live in
features.in/*/live/*scripts.d/ while generic for stage2/* in
features.in/*/stage2/*scripts.d/

(these *really* should be added atomically with the changes!)

This was tackled before but it took raorn@'s hint regarding
sys-freedos-linux (ms-sys has no bootsectors compatible with
freedos currently) and gns@' quick rush at make-freedos-floppy
script to wrap things up.

Should be further developed to become actually useful though.

wm-select should only be needed when multiple window managers
are available.

Should support ACPI and APM schemes;
only the former tested so far.

It was clear that "common" isn't very apt for packages that
will get *everywhere*, and became apparent when the need for
a "base+live packages" variable arrived with powerbutton feature.

So:
- the former COMMON_PACKAGES are now SYSTEM_PACKAGES;
- COMMON_PACKAGES act as "BASE+LIVE_PACKAGES".

Note that SYSTEM_PACKAGES also got factored out from stage2 based
features into stage2 subprofile itself; cleanups were due as well.

Rather minor fixups for things changed in the meanwhile and not
yet (re)documented properly; and a change for memtest feature
to require syslinux feature (the code's been changed to fit
the updated description, actually, and the change is purely
formal as no syslinux alternative is being used/planned so far).

distcfg.mk is nice for tracing the variable values' build-up,
but it's more useful to have a look at the final result sometimes.
So here it is, logged for each build and callable by hand.

It turned out that ~/out masked the broken execution path
when it's absent (which was handled incompletely).  Sigh.

MAIN_GROUPS should align better along with MAIN_PACKAGES
and MAIN_LISTS (even if MAIN_ prefix might be suboptimal
given that these packages are essentially extras within
the particular image).

minicd target was a hasty hack indeed, let's put it in line;
an attempt at practical use discovered the lack of important
packages like dhcpcd, these were added on per package basis
and quickly remade to use tags -- which seem to fit better.

It's still very immature -- base lists should be really base,
and functionality should be sorted out in more consistent way.
So far moved to tagged lists (which need some more experimentation
anyways to get both lists and their use more elegant).

- fixed live.iso (now actually useful):
  + extra cleanup was being done (coming from install2 case)
  + root user was password blocked
  + there was no unprivileged user (added "altlinux" w/o password)
  + added xdm setup hook for future X-based livecd flavours
  + several picks for a less slim "base" list
  + xdm login
- tweaked rescue.iso (added ext3grep)

Implementation based on m-p-d::profiles/scripts.d/03-syslinux
(but heavily modified for m-p of course)

- better *_LISTS printout
- somewhat cleaner .base comments
- more comprehensible git log messages
  + special handling for complex subprofiles
  + no need to commit build.log ;-)

dst a shell variable in a makefile recipe,
its dereference must be escaped properly;
the net result was false positive for the test
with no harm being done but extra stuff tossed
into the generated profile

These were broken since 5ff375aa10bee7937db457d3df3115f938bb6c77
due to no bootloader being tossed anymore by the very
installer-distro-altlinux-generic (which is the default)

squashcfk.mk is not to be attempted to copy unconditionally,
if we've got no kernel for stage2 there's no data (and hook)
to create it there

- introduced generic stage2 subprofile (non-standalone)
- ported installer and rescue over to stage2/{install2,rescue}
- initial stage2/live (needs more work for sure)
- use make-initrd-propagator
- updated and somewhat extended doc/

NB: mind #26133, #26134

NB: this implementation is awfully cut-n-paste one,
there are lots of common code with install2
which are subject to careful surgery

s,INSTALLER_KFLAVOUR,STAGE1_KFLAVOUR,g
s,INSTALLER_KMODULES,STAGE1_KMODULES,g

install2 isn't the only livecd around;
so far all of these share the same kernel
in m-p-d, and we'd rather need two+ kernels
with dual-arch media or so; let's not complicate
things too much at this point

make CLEAN=1 will prune workdirs after packing their results,
so disk (or preferably tmpfs) usage will be lower

NB: needs installer-distro-altlinux-generic = 6.0-alt2
    as it doesn't require alterator-grub indefinitely

This one is pre-alpha quality and very basic but complete enough
to crank out an icewm.iso successfully installing and booting
into xdm ;-)

(actually a week's on-and-off work including a pile of
underlying infrastructure stuff which was planned already
just not there yet)

I hereby release this work under the terms of GNU General Public License,
either version 2 or a later version at your option.

The code is either taken from mkimage-profiles-desktop.git (available
under GPLv2+) or written by me with some exceptions in the form of:
- snippets from GNU make documentation;
- stackoverflow.com and cmcrossroads.com articles;
- random advice on the 'net
to be strict; given the nature of such snippets being explicitly
proposed as an example or for immediate reuse, I consider it feasible
to redistribute those worked into this code.  If there's any concern
regarding copyright, please contact me by e-mail to sort it out.

In particular:
- .base is now generated from pieces (see image.in/Makefile)
- s/DISTROS/IMAGES/g; s/CONFIGS/DISTROS/g (for clarity)
- s/DISK_LISTS/MAIN_LISTS/g ("disk" was early m-p-d legacy)
- introduced BASE_PACKAGES to complement BASE_LISTS
- minor tweaks to Makefile (ARCH and DATE moved elsewhere)
- libdistro.mk: dropped overlooked IMAGE_INIT_LIST copy
- clean.mk: silly cleanup

installer-distro-server-light (using installer-feature-vm-server-light)
is specifically tuned for Server Light needs, and in general we have
different ones (regarding installer steps, autopartitioning profile,
configured services, etc).  So here's the replacement.

reword 'em, structurize 'em, facelift 'em too

This feature should absorb ha^Wmeasures for keeping install2
memory footprint to minimum, and so far it drops things not
critical for ru/uk/be users on the floor.

Should be ported to a l10n support framework when there's one.

We have several categories of overhead data:
- legacy 8-bit fonts, locales and gconv modules;
- illegible fonts (e.g. 5x7 or 6x10);
- obsolete vendor specific stuff;
- CJK (which would require proper support anyways).

Notes:
- JIS and things like VISCII are 8-bit either
  and thus might be reconsidered as "legacy";
- proper CJK support would probably include
  scalable fonts and an input method helper;
- main target audience is rather ru/uk/be by now.

(dropped wireless from install2 either)

A bit earlier the situation where there was a stage1 fallback
for INSTALLER_KFLAVOUR as the last of KFLAVOURS, so we'd have
an installer kernel or build would break; now the situration
got somewhat twisted: there could be a vmlinuz in stage1 but
no corresponding modules in install2 -- which can lead to
different surprises but at least alterator-vm would complain
about "Operation not permitted" on partition layout commit.

The fallback is a copypaste from sub.in/stage1/Makefile though
and should be redone properly somewhere.  The question so far
is, where exactly?