Items which can be updated (but have not yet been downloaded or queued for download)
will act as if the user selected "Ignore this update" from the app details view.
Items which represent app statuses (e.g. downloading, downloaded ready to install,
installed ready to run) will have the status removed. If required, we will also
forget that they are ready to install, so they wont be there next time.
Used to work, then the default join from `fdroid_app` to `fdroid_apk`
was removed for performance reasons. This adds the join back, but requires
queries to explicitly opt in to the join if they require it. The specific
query for known vulns is not a performance problem, because sqlite is able
to narrow the result set quite substantially before requiring a join onto the
fdroid_apk table anyway (e.g. by using the "installed app" table).
Untested because there are no apps in current repos which exhibit this
behaviour which have newer versions. Right now I'm testing with com.waze
from testy.at.or.at which only has the one version.
I'm also unsure of how important this is seeing as most the time it will
prompt people to update anyway.
Note that I don't think the query will work correctly across multiple repos,
because it is currently only querying the app with the "preferred
metadata".
Also use a newer version of testy.at.or.at index for the index-v1 test,
because it includes apks with "KnownVuln" anti features whereas the
older version did not.
When explicitly given an fdroidrepo(s) intent, it seems silly to restrict it based on
a path of /fdroid/repo, because it is plainly obvious it is an F-Droid repo.
Manifest and NewRepoConfig both had to be amended to allow this behaviour.
Fixes#1171.
There is a specific POSIX error "EIO" which seems to be the "general
purpose we don't know what went wrong but its probably bad" exception.
Our investigations in #855 resulted in the conclusion that it is likely
due to some sort of filesystem corruption or something like that.
Either way, it is annoying many people, so we need to prevent it or
ignore it, rather than prompting the user to submit a bug report.
After much investigation it was unable to be reproduced other than by
one of the original bug reporters. As such, this change ignores it.
Unfortunately Java `IOException`s don't have an API for getting the
errno of a POSIX IO error. Thus, this change results to parsing the
exception message instead :(
Fixes#855.
This improves performance when we need to decide whether or not apps are
installed or not while scrolling through large lists.
Fixes#1143.
Also change Jackson tests to properly ignore App#isApk.
This join resulted in one row for each apk in the result set (before
doing a GROUP BY), instead of one row for each apk. That is a large
difference in number of rows and resulted in much more work for sqlite.
Turns out this join wasn't required.
Some queries are deferred, and then forced to run by Android by invoking
`getCount()`. Under these circumstances, the measured speed of the query
execution is 1ms.
This adds speed logging around `getCount()` in case that is the first time
the query is run.
clbin is just for making the logs easy to read, if it fails, the build
should not be marked as failed, especially since the logs are probably
not needed if the rest of the job succeeded.
This is in the spirit of the setting, where users can see which apks are
available even though they are not installable.
Adds a message explaining why it is incompatible (i.e. because the
signature doesn't match the installed version).
