If Autofill shows a warning about an app whose publisher changed and the
user decides to trust the app and clear previous matches, they should
immediately be given the option to select a new match.
Previously, as AutofillPublisherChangedActivity did not return a
result, the old FillResponse with just a warning would be reused. We
now pass a useful response with no matches on to the activity, which
returns it after the user has chosen to reset the publisher info.
This is going to be less work to maintain going forward since we won't need to
change it every time we include a new subproject.
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
The newest versions of AndroidX Activity and Fragments correctly enforce the
requirement for all contracts to be registered at class init or before
the lifecycle has reached `Lifecycle.State.STARTED`. To comply with these requirements,
move all instances of `registerForActivityResult` being called at arbitrary points
in the code to be done at class init.
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
Filling passwords is usually only offered for fields with a password
input type, which ensures that they are always properly masked. Certain
custom views (e.g., the Termux terminal view), may apply their own
masking that is not based on the standard views.
With this commit, we allow filling passwords into arbitrary fields as
long as they have an Autofill hint that indicates a password.
Tie breakers for Autofill fields should only be evaluated if there is
a need to, in particular not if only a single or no field is matched.
Apart from a potential minor performance improvement, this should not
cause any user-visible changes, but does simplify the log output
considerably.
* BaseGitActivity: unwrap root cause for InvalidRemoteException as well
JGit's InvalidRemoteException, like TransportException, swallows more
useful errors as is clear from this (redacted) snippet.
```
D org.eclipse.jgit.api.errors.InvalidRemoteException: Invalid remote: origin
D at org.eclipse.jgit.api.FetchCommand.call(FetchCommand.java:26)
D at org.eclipse.jgit.api.PullCommand.call(PullCommand.java:41)
D at com.zeapo.pwdstore.git.GitCommandExecutor$execute$2$result$1.invokeSuspend(GitCommandExecutor.kt:2)
D at kotlin.coroutines.jvm.internal.BaseContinuationImpl.resumeWith(ContinuationImpl.kt:3)
D at kotlinx.coroutines.DispatchedTask.run(DispatchedTask.kt:15)
D at kotlinx.coroutines.scheduling.CoroutineScheduler.runSafely(CoroutineScheduler.kt:1)
D at kotlinx.coroutines.scheduling.CoroutineScheduler$Worker.run(CoroutineScheduler.kt:13)
D Caused by: org.eclipse.jgit.errors.NoRemoteRepositoryException: ssh://msfjarvis@[fe80::dead:beef]:22/pass-repo: fatal: '/pass-repo'
does not appear to be a git repository
D at org.eclipse.jgit.transport.TransportGitSsh.cleanNotFound(TransportGitSsh.java:14)
D at org.eclipse.jgit.transport.TransportGitSsh$SshFetchConnection.<init>(TransportGitSsh.java:20)
D at org.eclipse.jgit.transport.TransportGitSsh.openFetch(TransportGitSsh.java:1)
D at org.eclipse.jgit.transport.FetchProcess.executeImp(FetchProcess.java:1)
D at org.eclipse.jgit.transport.Transport.fetch(Transport.java:20)
D at org.eclipse.jgit.api.FetchCommand.call(FetchCommand.java:18)
```
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
* Add changelog entry
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
We don't use the new annotation added in the alpha and since this dependency is shared between the library and app it's best to stick to stable until necessary
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
Our tests have zero differences between these flavors so running 4 jobs rather than 2 just slows us down
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
* Ensure we're creating dialogs on the main thread
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
* Remove unused operation type
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
* Refactor launchGitOperation to use an enum
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
