Martin Matějek authored 4 years ago and Karel Koci committed 4 years ago

This adds call to sentinel-reload when sentinel configuration is changed
from Re/Foris.

Proper release without patch.

This adds call to sentinel-reload when sentinel configuration is changed
from Foris.

This is work in progress patch but we need this as soon as possible to
improve user's experience. Next fixup version should contain this change
and we should be able to drop this patch.

We are overwriting defaults. It is wrong as we didn't count with
variable ($PKG_VERSION). This means that if we didn't clean up
build_dir or dl folder, it will be extraced to smth without versioning.

- remove PKG_SOURCE_SUBDIR:=$(PKG_NAME) as well
- Add PKG_RELEASE if it was possible

This will save a lot of space on our buildbots

Drop restart of sentinel components

Also fosquitto package indentation was improved.

This separates foris-controller, reForis and Foris to separate
directories to not have single hude one.

Karel Koci authored 5 years ago and Karel Koci committed 5 years ago

This uses correct python install sections instead of replacing it with
its own. The problem was that python-package.mk does a lot of "standard"
fixes in that install and replacing it caused those fixes to be skipped.
There might have been originally reason on why to not use default
install but I don't see it now.

This also cleans a lot of other stuff around. Because we now use
python-package.mk provided install we use Py3Package install instead.
The same paradigm is used for optional ForisControllerModule install
section.

This also drops unnecessary hash-bang at the beginning of all scripts.
Those scripts are not run directly but instead sourced. It is bad
practice to do so because those scripts do not include functions they
might use (such as config_load and so on). Also note that in default
script is interpreted as shell so it is all right even if it is run
directly. This is just more cleaner solution.

This also drops all trace sets from same scripts. Most of those were
only calling service restart and that is pretty much useless to print
trace for that. There is only one exception and that is postinst of
controller-netboot-module. It is also doing UCI manipulation. This is
invalid in its core because it disallows configuration to be used. I am
going to move that change to separate oneshot package used to migrate
that configuration in future. That means that those trace outputs should
not be required anywhere.

There is one hack introduced here. Build system is checking if it is not
invoking some random Makefile by grepping it. In case of package they
look for pattern: call BuildPackage. This commit moves package
declaration from package it self to definition in included file. That
means that build system would ignore such Makefile. This introduces
comment with content matching pattern it looks for so it is not ignored.

There is a lot of foris specific packages now. Moving them to dedicated
directory cleans whole packages list and seems to be common approach in
upstream.

This also tries to fix netboot makefiles.