Before, `fdroid build` would create some subdirs, then fail when it could
not read the metadata via metadata.read_metadata(). This checks before
whether there is any metadata available, and warns the user accordingly.
Using the same JDK throughout should prevent weird bugs where a setup might
use Java8's jarsigner and Java7's keytool. This also allows the user to
set java_paths and have jarsigner and keytool used from that specified JDK.
This incorporates almost all of the patch that is in the Debian package
that forces fdroidserver to use the default JDK on that Debian release.
closes#93https://gitlab.com/fdroid/fdroidserver/issues/93
If we don't find the version name neither in the AndroidManifest.xml nor
in the build.grade file, we fallback to the tag the user specified to
search for.
That way we have a better version name than 'Unknown'.
I was wrong - we don't check for trailing whitespaces on lines. That
would have to happen at parse time, not in the linter, so it would slow
things down and would become an error, not a warning. Not really worth
it.
1. It is packaged in modern versions of distros, update docs accordingly
2. 1.1 is hopelessly outdated - support 1.4 onwards
3. Check the version properly, and write a better config (previously it
wrote the 1.1 config for 1.1 and 1.3+
4. Ensure that the default share from later versions is not present when
building, it's only required for provisioning.
Since we only sorted by count, ignoring the string, it meant that items
with the same count might be arranged in different manners. Hence the
`stats` behaviour was not predictable at all. Now it sorts first by
count, then by string.
It is a good idea in theory, especially for extra tasks specified by the
user. But in practice, tasks specified by developers often don't have a
clean counterpart. So this just breaks the build.
We were passing the utf-8 encoded string to textwrap, which took the
bytes as characters. Hence multi-byte unicode characters (in utf-8)
would count as multiple columns, which is clearly wrong.
This breaks on newer plugin versions. Instead of complicating this logic
even further to support them, upstream devs should instead use gradle
configs that actually work.
If their build fails because of lint, it's theirs to fix, not ours.
This allows us to run checks and fix types only on those that were
really modified. On our fdroiddata repo, with 2k apps with many unset
fields and flags, this reduces readmeta runtime from ~1.3s to ~1.1s.
