methods such as SQLiteStatement.simpleQueryForString() expect
one and only row to be returned when the query is executed.
if the application provides a query that doesn't return any data,
then the error message printed is pretty confusing.
make it less confusing and print the query itself to help debug
the error.
Change-Id: Ife2066f3a3eab0b98845a49e8f72b518458a7757
...to throttle contentobserver-based requeries
Why yes, I guess it could.
This also reworks AsyncTaskLoader to not generate multiple
concurrent tasks if it is getting change notifications before
the last background task is complete.
And removes some of the old APIs that had been deprecated but
need to be gone for final release.
And fixes a few little problems with applying the wrong theme
in system code.
Change-Id: Ic7a665b666d0fb9d348e5f23595532191065884f
Provide an API to allow application to support downgrade in databases
managed with SQLiteOpenHelper. Since onUpgrade() is now called only
if requested version is newer than current one, this adds support for
a similar onDowngrade() method, so customers can implement graceful
downgrading. If no downgrade method is implemented by caller, this
fallback to current behavior by throwing an exception.
Bug: 1609474
Change-Id: I0e0166984d4a700b1c43ce780b91244bc2fc87a6
bug:3083665
if sqlite can't open the database, delete it and create a new one.
what else is sqlite supposed to do?
Change-Id: I8b7c3b051fd1a146e4026f8f8acfdd099f9a70cf
the reason for bug:3281533, bug:3127159 is probably too many cursors are left
un-closed in the process.
print the info on the number of cursors left open when the exception
"cursorwindow allocation failed" occurs.
This should help us figure out if that indeed is the reason
and which process is leaving the cursors open.
Change-Id: I4b46be63f5dfbe9b102ad7a9cf9dd21e70f71e14
let cursor window size be set per device in device resources file.
default is 1MB.
for SR, it is 2MB.
it can be set to any value (in kB) in the device resource
strings.xml file
Change-Id: I67b1d04a5c9fc18b0cd4da6184d0b814b64d89e9
The current error message says "close failed due to unfinalised statements".
This CL fixes that message by incorporating one of the un-closed sql statements
should help developers debug the problem.
Change-Id: I2a24c6ba876caa008507b236052c5f03e8cbb27e
when downloads fail/get stuck, we need to look at the database state
for those downloads. and when the users report such problems, it is
a royal pain not to have that info and most users don't seem to bother
sending database dumps because it is a bit of work.
so lets just dump info about downloads that failed or
downloads from GSF (OTAs, for example)
helps debugging. there is STOP ship comment to not dump data once
HC is released.
Change-Id: Id1254982fd82b4c55f1816a2491f00966840f024
The signatures of the existing buildQuery and buildUnionSubQuery methods include a selectionArgs
parameter that is not actually being used in the method implementations. This parameter leads
to the misconception that SQL paramter substitution is carried out by these methods. I added
new variants of these methods without that parameter and deprecated the old variants.
Change-Id: I1bf770d5c777649e9aac36d93aa93bd65bbcc2a3
This warning doesn't really matter much. If a developer wants to
debug the cache usage, it can be done by turning on this flag.
adb bugreport also displays cachestats - which can be used as a debugging tool.
Change-Id: Ied173714d535c271133247ee4768f86d3be359cf
the warning printed currently "do requery on background thread"
is not that useful in processes such as gapps, acore.
to reduce the deluge of bugs assigned to me, I think a simple
deprecation warning is better.
Change-Id: I7a1129ea889f10e72895092a3cdd48cc96d0d1f0
Make clear in the Javadoc comments of the `Cursor` get* methods that
implementations thereof can have implementation-defined behavior. In some cases,
these changes actually correct the documentation. For example, in the case of
`getShort` and the `SQLiteCursor` implementation thereof, non-numeric data is
*not* converted to a `short` via Short#valueOf or even in a functionally-
equivalent manner.
Change-Id: Ib2f81811a603680b52fc482eb9c0f3195447566f
In two places involving locking, reordered the code so that the lock
acquisition is performed outside of the `try` block and everything
else that needs to run while the lock is locked *within* the `try`
block.
Change-Id: I3dad2c4bbf60b219fc6db2aa35e2ed296cb39128
database lock() should not be called from a synchronized methods
in database layer. add code to check.
this will catch potential deadlock situations sooner
Change-Id: I9d913f5e2af304f4d9ad5b2641c3a768c4bc97f9
This makes it more future-proof and maintainable, not exposing the
internal bitpacking state.
The implementation is unchanged (the policy is still just an int we pass
around).
Also starts to introduce VmPolicy, for things which are process-wide,
not per-thread. As an initial user, make SQLite's Cursor finalization
leak warnings use StrictMode.
Change-Id: Idedfba4e965716f5089a52036421460b1f383725
Synchronized blocks should never call lock() or any operation that
does it. otherwise, bugs like bug:3032897 will occur.
Remove synchronization on all methods in SQLiteStatement
and SQLiteProgram because the caller is supposed to
do the synchronization if the same SQLiteStatement or
SQLiteQuery (which extends SQLiteProgram) object
is used by more than one thread at the same time.
(documentation on SQLiteStatement, SQLiteProgram
mentions this note)
Change-Id: Ifd6ee94d743c38c187e2eb778c57076da7a15073
