Don't remove the container when `DEBUG` variable is set

* And get rid of the `sleep infinity` hack instead; one can simply start
  the container again
* Document the workflow for instigating build failures

README.md

@@ -103,6 +103,12 @@ edit `~/.config/containers/storage.conf` (or `/etc/containers/storage.conf` for
 system-wide configuration) to set `runroot` and `graphroot` to different
 locations.
 
+### Investigation of build failures
+By default, `makecontainerpkg` starts the container via `--rm` so the container
+is removed in the end. Set `DEBUG=1` to prevent that. Then one can use e.g.
+`podman container start …` and `podman container exec -it … bash` to enter the
+container for manual investigation.
+
 ---
 
 There's also the 3rdparty repository

devel/container/makecontainerpkg

@@ -2,7 +2,8 @@
 set -e
 
 # parse arguments
-cre_args=(--workdir "/startdir" -v "$PWD":/startdir --rm)
+cre_args=(--workdir "/startdir" -v "$PWD":/startdir)
+[[ $DEBUG ]] || script_args+=(--rm)
 script_args= read_script_args= no_sync=
 for arg in "$@"; do
     if [[ $read_script_args ]]; then
@@ -26,9 +27,4 @@ if ! [[ $no_sync ]]; then
     script_sync=$(cat "$bindir/containersync")
 fi
 
-# allow one to prevent the container from stopping via DEBUG variable
-if [[ $DEBUG ]]; then
-    script_args+=' ; sleep infinity'
-fi
-
 ${CRE:-docker} run "${cre_args[@]}" "${CRE_IMAGE:-archlinux-base-devel}" bash -c "$script_sync $script $script_args"