This commit includes following changes
* add newly added option to configuration referenec
* remove few legacy options from configuration schemas used for
validation, which might lead to errors during validation.
Note, however, that settings will be still read by the service
* add link to aurcache
* hide service-setup command description under spoiler
It has been found that if there are missing dependencies than whole
process will break instead of just skipping packages. During package
addition it is fine-ish, but it will break updates run
it has been reported that duriing reading pkgbuilds with latin-1 charset
the execption will be raised. Well, it is one more point to rewrite
parser to use own impl instead of shlex and parse raw byte array instead
Previously if package is VCS and version in PKGBUILD doesn't match to
AUR one, then makepkg will update pkgbuild ignoring all previous pkgrel
patches
With this change during task init dry ryn process is always run for vcs
packages
Old behaviour leads to variable to be serialized always in single quotes
(if shlex decides to quote it). It doesn't allow to expand variables during execution,
so those values should be put inside double quotes instead. This commit
checks if there is sign char in the variable and quotes manually,
otherwise returning shlex call
If value is stored as array in database it is serialized as json, but
read as normal string, which lead to innability to use list patches
This fix also removes any postprocessing (unquoting) for functions
There are two major changes here. First of all, the image generation now
consist of two separated stages, the build itself and the production
image generation. Secondly, the packages inside image are now installed
as they were at the time of the root image generation (defined by stat
command)
Another side change is that container does not longer ship syncronized
(and out-of-dated) pacman databases; they have to be synced manually
passlib uses deprecated crypt module which is deprecated and scheduled
for removal in 3.13. Unfortunately, this module seems to be
unmaintained, so this commit replaces passlib with bcrypt, unfortunately
breaking current passwords
