It's a preliminary change for IMAP bug fixes.
Also,
- Fixed a potential bug in ImapFolder.setFlags where it'd throw
StringIndexOutOfBoundsException if flags is empty.
- Added a generic flag to proguard.flags so that now all methods with
the "ForTest" sufix are automatically preserved.
Turned out it wasn't needed for this CL, but it should come in handy
someday.
Bug 2538076
Change-Id: I49a08afc196c7b7f1f30477dfc38ac5381045d84
* Yahoo! is not supporting search by UID so I can't identify the new
the UID after I upload. This inability to correlate the local and
remote messages means that we wind up syncing the same message back
down, in a loop, which spawns more messages.
* Yahoo! has partial support for UIDPLUS, and reports the new UID when
I append (upload) messages.
* Modify IMAP parser to parse response lists
* When APPENDUID is reported, use it (and skip the search)
* Modify the few other existing users of response lists to use the
parsed versions instead. Provided a couple of lightweight utilities
to make it easier to work with ImapList.
* Unit tests for most of it.
* Optimization: share a static date/time parser for all IMAP connections
Bug: 2448220
Change-Id: Ic10fc1a195ccf4671a498188cc8b17848c8d9df7
Added DiscourseLogger, which stores last N (currently 64) lines of IMAP
commands sent to the server and responses received from the server.
We dump it to logcat when the IMAP parser crashes, that is, a) getting a
RuntimeException in ImapFolder.fetch() or b) getting a Runtime/IOException
in ImapResponseParser.
Bug 2480227
Change-Id: I6b5a728a7df106627ec29bb3c7c04a97a99b444b
Message and MimeMessage were creating a lot of unnecessary sub-objects
even when not needed, so do a bunch of lazy initialization. This should
raise the bar on the size of gigantic inboxes giving us trouble.
* Specific optimizations:
* Replace date formatter with a shared static
* lazy create mHeader (ArrayList)
* lazy create mFlags (HashSet)
* optimize MimeHeader fields class
* lazy create local message-ID (expensive-to-make uuid String)
* make message-id string less expensive to create
* Other cleanups:
* add some override annotations
* privatize some members
* update a fragile test (not a deep fix, it's still fragile)
Side effect, should be faster too.
Bug: 2357564
Bug: 2093422
Change-Id: I8a873879d402e2662339d5398ad0b15da6e580e9
We've observed that the secure.emailsrvr.com email server returns an excess
FETCH response for a UID FETCH command. Excess responses don't have the
UID field, even though we request, which led the response parser to crash.
This patch fixes it by making the parser ignore response lines that don't
have UID.
Bug: 2441065
- The entire android.pim package is hidden.
Use java.text.ParseException instead of android.pim.DateException.
- TelephonyManager.getDefault() is hidden.
Use Context.getSystemService() instead.
- Use newly added Base64 in the framework.
Bug 2226160
Convert all usages of com.android.email.codec.binary.Base64 to use
com.android.common.Base64 instead, except for Base64OutputStream
(which doesn't exist in android-common yet).
Change-Id: I339a1f451245138187080c7444fcabef8d13f8aa
* scrub all external strings to keep them compliant for IMAP protocol
* move Build.MODEL to x-android-device-model
* send x-android-mobile-net-operator
* send AGUID
* unit tests for above
* retrieve providers from VendorPolicyLoader
Bug: 2332183
* Add IMAP ID command to all login sequences
* Send generic information for now
* Explicitly catch & discard parsing errors, since we really don't
care if the command succeeds or not.
* Unit tests
Bug: 2332183
Some IMAP servers return NIL if you BODY.PEEK[TEXT] a messsage with
no body, instead of the more canonical {0}CRLF. Instead of messing with the
parser to deal with that, it makes more sense not to try and fetch empty
bodies. So there are three changes:
* Don't fetch parts when size = 0
* Don't append "null" when there is null body text
* Slight change to attachment handling so size is reported >0
* Unit tests on some of the related lower-level protocol stuff
Bug http://b/issue?id=2160387
Change-Id: Ifb8fb0ed5ce7297908e1ae8d5a02dda5975c4a3c
* Add "Accept all certificates" modes to incoming/outgoing secure choices
* Change URI scheme slightly to make "trust" a flag, not part of the
protocol.
* Change Stores to know about new URI scheme
* Slightly rework Transport API to make "trust" an independent flag
* Adapt HostAuth to handle new Uri scheme
* Remove the old ambiguous "optional" code, which was allowing
some unsigned certificates, but was *also* allowing TLS to
optionally start (though not SSL, despite the UI strings.)
* Add a few unit tests to EmailContent
* Add logging and a bunch of comments to TrustManagerFactory, and a bit
of simple cleanup to make it more readable.
* Add missing conversion of SSLException->CertificateValidationException
in TLS so we get the correct certificate errors from TLS too.
* Re-enable TLS for mac.com accounts (which had a certificate problem)
Fixes bug http://b/2119755, http://b/1374780, and probably a raft of
earlier and/or external bugs about certificate problems.
Change-Id: Iaf99a8da3eaadaa4cdeec224737838b5d6813e55
I don't know the root cause of the null pointer (possibly a broken
connection earlier in the sync) but we shouldn't be crashing here.
Fixes http://b/2135743
* Create logic to detect upsyncable messages in Sent
* Note: Drafts is now local only for IMAP - no sync, either way
* Rewrite MessageController.processPendingAppend for Provider world
* Write provider message -> legacy message converter
* Fixed bug in IMAP APPEND (it was not picking the right UID for the
uploaded message.)
* Better handling of server internaldate
* Add constants for new X-Android-Body-Quoted-Part header
* Add EmailContent routines to get each of the 5 parts of the body
* Remove "Load more" from unsynced message lists
* Add toString to MimeHeader for debug support
Bug # 2097471
TODO (next CL): Upload attachments records too
Change-Id: I209182f5adc6b6696919f559e3cbbdd58b3eed3a
* Add \FLAGGED support to IMAP (writeback)
* Add code in Controller to kick MessagingController
* Rewrite pending commands system to scan through provider's updated
messages table and react
* Fix a unit test that I broke
* Cleaned out some of the old PendingCommand support
Addresses the 2nd half (upsync) of bug 1904385
TODO:
Can I add a unit test for IMAP flag writer?
Change-Id: I5a96a695d4f35fca1395506f165b86d9fb19b543
* Set the star and the read/unread states properly when a
message is downloaded for the first time.
* Update them on already-downloaded messages.
This is download only - not upload
Bug 1904385
Change-Id: Id03a0957677bb39f4a57ed0542eaa8accc36ab48
* Handle messages >25k
* When structure is available (e.g. IMAP) pull in the entire body
and the list of attachments
* When structure is not available (e.g. POP) pull in a large chunk of
the body to try and capture the message body at least.
* Implement loadAttachment for IMAP/POP to demand download large items
* Tested with IMAP & POP messages
INCOMPLETE (file bugs):
* implement logic for the old loadMessageForView calls that comes from
MessageView (when you open a message that's partially-loaded)
* Resolve handling of mimetype when attachment info is read (currently
we're assuming base64 in a couple of places)
* delete account => delete attachments
* delete attachment => delete file
* create account => clear existing attachments for acct id
due to API change, but still has a smaller footprint. Also fixes the
bug in the original, which is that we actually needed to udpate the
local trash folder, not the remote one.
BUG=1807499
Automated import of CL 147714
*** Reason for rollback ***
We figured out a simpler solution affecting fewer files - we
don't actually need the new remotestore API.
*** Original change description ***
Some stores require changing the UID of a message when it is
copied to a new folder (I'm looking at you, EAS). Add a callback
to Folder.copyMessages() which allows the store to report back
such changes. Then, add a new api to record the new values:
Folder.updateMessages().
For now, the two APIs are linked by a common callsite in
MessagingController, so the existing stores can use a minimal
implementation - if they don't call the callback, nobody will
call the update.
BUG=1807499
Automated import of CL 147708
copied to a new folder (I'm looking at you, EAS). Add a callback
to Folder.copyMessages() which allows the store to report back
such changes. Then, add a new api to record the new values:
Folder.updateMessages().
For now, the two APIs are linked by a common callsite in
MessagingController, so the existing stores can use a minimal
implementation - if they don't call the callback, nobody will
call the update.
BUG=1807499
Automated import of CL 147620
syncing. This provides a key-value store, per folder, that
can be used by network Stores to record persistent data such
as sync status, server keys, etc.
Note that, by definition, this only applies to remote folders
(e.g. IMAP, POP3). You'll see everywhere that LocalFolder is
passed null, and this is correct - LocalFolder *is* persistent
storage and does not need external help.
Note to reviewers: The core changes are Folder.java,
LocalStore.java, and LocalStoreUnitTests.java, so please give
them the bulk of your reviewer attention. The other files
are just following along with minor API changes. Of those,
the one worth close examination is MessagingController.java,
which is the only place in the system where remote Folders
are bonded with Local Folders and thus where this new API
comes into play.
Note to jham: Can you please take a look at
LocalStore.LocalFolder.setPersistentString() and recommend
better SQL foo than my primitive test-then-update-or-insert
logic, which is not transactional or threadsafe.
BUG=1786939
Automated import of CL 146134
The current design for Store classes (e.g. IMAP) did not provide for
any persistent storage. This is the beginning of a mechanism to
provide that. It's quite simplisitic - each Store can read/write one
persistent string - but that's enough for the first simple use case
(saving some sync data for EAS).
The core changes here - suggest reviewing first - are in Account.java,
Store.java, and AccountUnitTests.java. Everything else is just
following the API change that was necessary.
Note that, by definition, this only applies to remote stores (e.g.
IMAP, POP3). You'll see everywhere that LocalStore is passed null, and
this is correct - LocalStore *is* persistent storage and does not need
access (so far, at least).
BUG=1786939
Automated import of CL 146061
Makoto Onuki