more preventing of "apt clean"

[umlaeute]

While b47534c provides a PKGUPGRADE_CLEANUP variable to prevent the pkgupgrade module from running apt-get clean in the cleanpu script, the base module still unconditionally clears the apt-cache:

CustomPiOS/src/modules/base/end_chroot_script

Lines 41 to 42 in 8a3a002

	#cleanup
	apt-get clean

This PR introduces a new BASE_APT_CLEAN variable, that can be set to anything but the default yes to prevent apt-get clean from being run.

while I edited the files i couldn't resist the urge to drop some unnecessary bashisms and replace the oldstyle command substitution `` with the modern, stackable $() (but these changes are in separate commits, so feel free to ignore them)

[umlaeute]

i wonder whether the PKGUPGRADE_CLEANUP flag should somehow inherit the BASE_APT_CLEAN value (unless set by the user).

but that's mostly a design choice, how independent modules must be (and whether the base module is an exception)

[guysoft]

Thanks for fixing the square brackets for string comparison

[guysoft]

@umlaeute
This seems to break now for me, not sure why, investigating

[umlaeute]

hmm. i obviously did not test in docker, though i do not understand why it would break.

the only possible breakage i can think of is, that the pushd/popd solution handles the case gracefully where ${CUSTOM_PI_OS_PATH} does not exist at all (in which case you could get the commitish of the current directory).
if this is indeed the desired behavior, then i think a somewhat easier to read solution would be

[ -n "$BASE_COMMIT" ] || BASE_COMMIT=$(cd "${CUSTOM_PI_OS_PATH}"; git rev-parse HEAD)

(since the $() (like the ``) is executed in a sub-shell, there's no need to go back to the calling directory)

this would fail (in the case of a non-existing ${CUSTOM_PI_OS_PATH}) only if the the current directory isn't git-tracked. (though i think in this case the original code would fail as well)

otoh, the original $(pushd /bla; git rev-parse HEAD; popd) (that is: with an non-existing ${CUSTOM_PI_OS_PATH}) returns an error as well, so the script would fail if this dir did not exist.

🤔

[umlaeute]

my reference to docker was triggered by #206 (comment) (for whatever reasons I assumed that the comment on is related to this regression)

[umlaeute]

could you provide an example (or link to a repository) that shows the problem?

[guysoft]

Related:
#209

[guysoft]

@umlaeute The issue derives because the git repo is not available from the docker container. The solution would be to keep the commit that the container was built from is a agreed location. Then search for that, then for the git repo.

[umlaeute]

i came to the same conclusion (but obviously forgot to comment).

and yes, i think that it should be part of the Docker buildout to embed such a frozen version information within the container (at least, if the BASE_COMMIT info is of any relevance with the Docker container; i guess mostly you would be interested in the commitish of the "consumer" OS, rather than the base CustomPiOS commitish)