and plymouth is in use. Previously, this would use plymouth's
ask-for-password action, passing it a --command argument that
included a pipeline (line 206 in previous version of 00-crypt.sh).
As plymouth does not shell-interpret its command however, this fails.
In this version, we create a temporary script file with the necessary
pipeline command in it, and pass this temporary script to plymouth instead.
The temporary script is deleted after use.
(In theory, it should be possible to construct a "/bin/sh -c 'cmd'"-style
invocation for the --command, thereby avoiding this, but I have had no
luck with this approach, due to plymouth's argument parser.)
After generating the module dependencies for drm drivers, we
should pick up all the modules by looking at the whole kernel
modules directory path and not just at drivers/drm.
This has been reported in GitHub issue #17 for genkernel-next.
systemd will not work, and setting it to /sbin/mdev doesn't work either
since 99% of the non-systemd systems use udev/eudev.
If you really want to set the hotplug handler, do it in the kernel
config.
Since we now have ldconfig which can run during the boot, make sure we
unconditionally copy all the other configuration files (ld.so.conf and
relatives), while only skipping the "cache" step.
A couple of problems here.
1. CONFIG_LOCALVERSION was never added the kerncache.config.
2. Parsing of CONFIG_LOCALVERSION expects kernel .config style syntax.
For this point I opted for the extra space, to make the
kerncache.config syntax consistent. There is no good reason to even
keep the CONFIG_LOCALVERSION name.
This looks like a leftover from a141d71531.
All occurrences of NO_KERNEL_SOURCES should have been replaced by
the negated KERNEL_SOURCES.
Though probably isTrue should be used for all boolean options, to avoid
confusion when setting them in the config file (some of them should be
set to 0 or 1, while others support yes/no/0/1/true/false).
docache support was broken since the Stone age. Not all the
loop types correctly supported it because it wasn't transparent
to them and required manual hackery. Now _setup_livecd_content
transparently configures LOOP_PATH if docache is enabled.
Prioritize the latter (init=) if it's set and no real_init= is set.
This, while keeping backward compatibility, makes possible to
directly use init= to boot the preferred init system.