Connection establishment now happens only in one place in each IMAP loop.
Now all connection establishment happens in one place
and is limited by the ratelimit.
Backoff was removed from fake_idle
as it does not establish connections anymore.
If connection fails, fake_idle will return an error.
We then drop the connection and get back to the beginning of IMAP
loop.
Backoff may be still nice to have to delay retries
in case of constant connection failures
so we don't immediately hit ratelimit if the network is unusable
and returns immediate error on each connection attempt
(e.g. ICMP network unreachable error),
but adding backoff for connection failures is out of scope for this change.
If a time value doesn't need to be sent to another host, saved to the db or otherwise used across
program restarts, a monotonically nondecreasing clock (`Instant`) should be used. But as `Instant`
may use `libc::clock_gettime(CLOCK_MONOTONIC)`, e.g. on Android, and does not advance while being in
deep sleep mode, get rid of `Instant` in favor of using `SystemTime`, but add `tools::Time` as an
alias for it with the appropriate comment so that it's clear why `Instant` isn't used in those
places and to protect from unwanted usages of `Instant` in the future. Also this can help to switch
to another clock impl if we find any.
This change depends on async-imap update that resets the timeout
every time an `* OK Still here` is received.
Reducing timeout allows to detect lost connections
not later than 6 minutes
because Delta Chat will attempt to finish IDLE with DONE
after 5 minutes without keepalives
and will either get TCP RST directly
or, worst case, wait another minute for TCP socket read timeout.
Ensure the client does not busy loop
skipping IDLE if UIDNEXT of the mailbox is higher than
the last seen UID plus 1, e.g. if the message
with UID=UIDNEXT-1 was deleted before we fetched it.
We do not fallback to UIDNEXT=1 anymore
if the STATUS command cannot determine UIDNEXT.
There are no known IMAP servers with broken STATUS so far.
This allows to guarantee that select_with_uidvalidity()
sets UIDNEXT for the mailbox and use it in fetch_new_messages()
to ensure that UIDNEXT always advances even
when there are no messages to fetch.
This prevents accidentally going IDLE
when the last new message has arrived
while the folder was closed.
For example, this happened in some tests:
1. INBOX is selected to fetch, move and delete messages.
2. One of the messages is deleted.
3. INBOX is closed to expunge the message.
4. A new message arrives.
5. INBOX is selected with (CONDSTORE) to sync flags.
6. Delta Chat goes into IDLE without downloading the new message.
To determine that a new message has arrived
we need to notice that UIDNEXT has advanced when selecting the folder.
However, some servers such as Winmail Pro Mail Server 5.1.0616
do not return UIDNEXT in response to SELECT command.
To avoid interdependencies with the code
SELECTing the folder and having to implement
STATUS fallback after each SELECT even when we
may not want to go IDLE due to interrupt or unsolicited EXISTS,
we simply call STATUS unconditionally before IDLE.
Gmail archives messages marked as `\Deleted` by default if those messages aren't in the Trash. But
if move them to the Trash instead, they will be auto-deleted in 30 days.
IMAP capabilities and selected folder are IMAP session,
not IMAP client property.
Moving most operations into IMAP session structure
removes the need to constantly check whether IMAP session exists
and reduces number of invalid states, e.g. when a folder is selected but
there is no connection.
Capabilities are determined immediately after logging in,
so there is no need for `capabilities_determined` flag anymore.
Capabilities of the server are always known if there is a session.
`should_reconnect` flag and `disconnect()` function are removed: we
drop the session on error. Even though RFC 3501 says that a client
SHOULD NOT close the connection without a LOGOUT, it is more reliable
to always just drop the connection, especially after an error.
Note that `IMAP IDLE protocol timed out` was previously
added to the wrong error: not the timeout error (first `?`)
but actual error happened during IDLE, such as a network error.
* Make set_config() look a bit nicer
* Add OnlyFetchMvbox option
* Add test for the config
* Add option to only watch mvbox
This is supposed to support having a server-side rule which moves
emails to the mvbox already. The new option makes sure the mvbox is
wathched and also makes sure no messages are feched from folders other
than the mvbox and the spam folder if enabled. It does not interact
with the other settings.
* Fixup ignore conditions
* Cleanup some bits
* Watch the mvbox when `WatchMvboxOnly` is set
* Rename back to only_fetch_mvbox (flub said it's OK for him)
* typo
* clippy, more typos
Co-authored-by: Hocuri <hocuri@gmx.de>
- Replace .ok_or_else() and .map_err() with anyhow::Context where possible.
- Use .context() to check Option for None when it's an error
- Resultify Chatlist.get_chat_id()
- Add useful .context() to some errors
- IMAP error handling cleanup
See https://support.delta.chat/t/discussion-how-to-show-error-states/1363/10 <!-- comment -->
It turns out that it's pretty easy to distinguish between lots of states (currently Error/NotConnected, Connecting…, Getting new messages… and Connected). What's not that easy is distinguishing between an actual error and no network, because if the server just doesn't respond, it could mean that we don't have network or that we are trying ipv6, but only ipv4 works.
**WRT debouncing:**
Sending of EVENT_CONNECTIVITY_CHANGED is not debounced, but emitted every time one of the 3 threads (Inbox, Mvbox and Sentbox) has a network error, starts fetching data, or is done fetching data.
This means that it is emitted:
- 9 times when dc_maybe_network() is called or we get network connection
- 12 times when we lose network connection
Some measurements: dc_get_connectivity() takes a little more than 1ms (in my measurements back in March), dc_get_connectivity_html() takes 10-20ms. This means that it's no immmediate problem to call them very often, might increase battery drain though. For the UI it may be a lot of work to update the title everytime; at least Android is smart enough to update the title only once.
Possible problems (we don't have to worry about them now I think):
- Due to the scan_folders feature, if the user has lots of folders, the state could be "Connecting..." for quite a long time, generally DC seemed a little unresponsive to me because it took so long for "Connecting..." to go away. Telegram has a state "Updating..." that sometimes comes after "Connecting...".
To be done in other PRs:
- Better handle the case that the password was changed on the server and authenticating fails, see https://github.com/deltachat/deltachat-core-rust/issues/1923 and https://github.com/deltachat/deltachat-core-rust/issues/1768
- maybe event debouncing (except for "Connected" connectivity events)
fix https://github.com/deltachat/deltachat-android/issues/1760
`Imap` structure is always created in a configured state now. There is
no default value for `ImapConfig` anymore.
Also resultify Scheduler::start() to fail on database errors, for
example if IMAP configuration cannot be read from the database during
`start_io()`. Previosuly errors during reading keys such as
`mvbox_watch` were simply ignored and folders were not watched until
the application is completely restarted, now start_io() will fail and
scheduler will only be started at the next start_io() call which
usually happens when app is brought to the foreground.
Using `impl AsRef<str>` as the argument instead of `&str` makes it
possible to call the function with `&str`, `String` and other types
that implement `AsRef` trait.
The cost of it is that compiled binary contains mulitple versions of
the same function, one for each variant of types. If function contains
multiple generic `impl AsRef` arguments, the number of versions possibly
compiled into binary grows exponentially with the number of arguments.
Simple way to avoid it is to call `.as_ref()` on the caller side to
convert the argument to `&str`. In most cases even adding a `&` and
relying on `Deref` coercion is sufficient.
This patch changes many functions that accepted `impl AsRef<str>` and
`impl AsRef<Path>` to accept `&str` and `&Path` instead.
In some places `.clone()` calls are removed. Calling `.clone()` on
`String` and passing `String` to a function accepting `impl
AsRef<str>` is completely unnecessary as `&str` reference could be
passed instead. There is no clippy warning against it yet, but
changing argument type to `&str` allowed to find these cases.
The result of debloating is not impressive, several hundred kilobytes
are saved, which is about 3% of the `.so` binary, but the code is
cleaner too.
This removes the proxy via crate::error to depend on anyhow directly.
There is no benefit to this indirection and this makes it simpler to
see which error types are used.
Start implementing #1994
TODO (in later PRs):
- Add a hint to the watch settings that all folders are fetched from time to time (to be done in the individual UIs)
- folder names are case-insensitive, so double-check that all comparisons are case-insensitive
- The `scan_folders.rs` file didn't get as large as I expected and it's probably not worth it having an extra file for it. But if there are no objections, I'll make another PR to rename it to `folders.rs` and also put into it `configure_folders()` from `imap/mod.rs` and `needs_move()` with all its tests from `message.rs`.
Done:
- Most mailboxes have a "Drafts" folder where constantly new emails appear but we don't actually want to show them, what do we do about this? The most reliable way to detect such messages that we found up to now is:
If there is no `Received` header AND it's not in the `ConfiguredSentbox`, then ignore the email.
- before or after INBOX idle trigger a new "scan all folders for messages". It does a "list folders" and then goes through all folders with select-statements, checking if "next-uid" was changed since checked last time. This might be batchable but in any case should not consume a lot of traffic. We might debounce this scan activity to happen at most every N minutes
- if next-uid changed for a folder, we "prefetch" and "fetch" DC-messages as as needed ("dc-messages" are not just those with "Chat-Version" headers, but can also be regular emails)
- if we discover DC-messages in folders that have the "/Spam" flag (maybe excluding ContactRequests) we automatically move them to INBOX/DeltaChat folder to help provider-spam-systems to regard this contact/mail as non-spam
- for now, we do not change any user visible option, but introduce this "scan all" automatically and on top of what exists. The DeltaChat folder-watching does not perform scan-all-folders (maybe with the exception to trigger scan-all also with DeltaChat if INBOX is not watched)
- Tests (except if you have ideas to improve them)
- all folders, their last uidvalidity, next-seen etc. are kept in a separate "imap-sync" sqlite table. Maybe this can be used to streamline some of the "Sent" folder and "DeltaChat" folder detection code we already have.
- We now also move self-sent messages from the Inbox to the Sent folder if `mvbox_move` is off, as this was very easy to do now. This way, we now behave more like a normal MUA if the user wants this.
FOR LATER PRs:
- maybe for the first 50 messages or so, we could reduce the IDLE-timeout (currently 23 minutes or so) to faster detect messages sent to non-inbox folders. However, on Android and iOS, we would likely trigger scan-all when the app moves to foreground, and so it might not be neccessary to reduce the current idle-timeout at least for them. We can leave this "faster discovery" question for the end, after we move to real-life testing.
- (Later on, after the above works, we can consider heuristics on which folders to perform IDLE on, and remove the Watch-folder options (inbox, deltachat, sent). We tried to find a safe scheme for already doing it but failed to fine one, too many unknowns, also some questions regarding multi-device (you might have different settings with each of it, one moves, the other doesn't etc.) so we postponed this in favor of the above incremental suggestion.)
* Start implementing #1994
* Add debug logs, it seems like the SQL migration can go into another pr
* Let fetch_new_messages return whether there are new emails
* Code style
* Don't prefetch if there are no new emails
* clippy
* Even more debug logs
* If the folder was not newly selected, return always try to fetch as
uid_next is probably outdated
* Fix new bug
* Recognize spam folder
* if we discover DC-messages in folders that have the "/Spam" flag (excluding ContactRequests) we automatically move them to INBOX/DeltaChat folder to help provider-spam-systems to regard this contact/mail as non-spam
* Clippy, prioritize folder_meaning over folder_name_meaning
* Add a first test, for the first day after installation only debounce to 2s
* Start adding two tests (both of them fail)
* Don't abort folder scan if one folder fails
* More consts
* Replace bool return value by enum
* Split test up into multiple tests
* Print logs during rust tests
* Rust tests pass now
* .
* One of the Python tests passes now - reconfigure folders during scanning
* Make the last test pass - Delete emails in all folders when starting the test, not only inbox and mvbox
The problem had been that emails were left in the folder "xyz"
* lint
* DB migration (untested)
* Store uid_next in SQL instead of lastseen in a config
* Revert "If Inbox-watch is disabled and enabled again, do not fetch emails from in between"
all folders are always watched, anyway
* clippy, rm debug logs, comments
* Codestyle, comments
* fixing things again
* Fix another test: don't fetch from uid_next-1 but uid_next; make some {} to {:#} so that we can use `.context(...)`
* move self-sent, non-setupmessage chat messages to the Sent folder if `mvbox_move` is off
* comment
* Comments, make sure things work even if there is no uid_next
* Style
* Comments
* The rust test tested wrongly
* comments, small codestyle change
* Ignore emails that are probably only drafts
Most mailboxes have a "Drafts" folder where constantly new emails appear
but we don't actually want to show them.
So: If there is no Received header AND it's not in the ConfiguredSentbox,
then ignore the email.
Also: Add test.
* Fix occasional test failure, it was introduced as DC now moves messages from Inbox to Sent
* Add `Received` header to the rust tests
* After this PR we will always watch all folders and delete messages there if server_delete is enabled. So, for people who have server_delete on, disable it and add a hint to the devicechat
* comment, small fix
* link2xt's first review
* Use ON CONFLICT(FOLDER) DO to update and if it doesn't exist, then insert
Reason from link2xt: We had a problem with multiple peerstates inserted due to key fingerprint parsing error previously. With logic in Rust a similar problem can occur: an UPDATE can fail for reasons other than a conflict. PRIMARY KEY should ensure uniqueness in this case, but anyway.
* Remove two TODO statements, remove fetch_new_messages: ignoring uid {}, uid_next was {} log
* Next TODO: Make uidvalidity and uid_next DEFAULT 0
* rm two TODOs, Seems like we are not going to `exclude folders that are watched anyway` in this PR
* small tweak: Handle instants more carefully
* Add scan_all_folders_debounce_secs config for tests, set debounce to 60s (before it was just 2s during the first day)
* Don't use bold letters for the device message
* React to changes in the folders better
Before, if there was a configured Sent folder, but then it got
removed and replaced with another folder with a name meaning "Sent" but
without Sent flag, it would be ignored.
So, instead of checking against ConfiguredSentboxFolder,
create two Option variables at the beginning of the loop and replace
them with Some if it is None. At the end of the loop, store the new
values into ConfiguredSendboxFolder and ConfiguredSpamFolder, even if it
is None.
Also, derive some useful traits.
* move job: Return a meaningful error if server_folder is None instead of panicing
* small error-handling fix
* Fix test_fetch_existing() python test
Before, we sometimes got a race condition where scan_folders() sees that
there is a Sentbox and saves this info after we set the
ConfiguredSentbox to None and before the message is sent.
So, just expect that the message is moved to the sentbox.
* migration is 72 now
* rm 2 TODOs, Don't infinitely retry when dc_receive_imf() returns Err
* clippy: Remove glob imports
* Delete created folders at the beginning of tests
(some created folders made problems in the next tests because)
* Improve resetting accounts between tests
Read all of an e-mail accounts messages and extract all To/CC addresses
if the From was from our own account.
Then, fetch existing messages from the server and show them.
Also, I fixed two other things:
- just by chance my test failed because of an completely unrelated bug.
The bug: bcc_self messages were not marked as read if mvbox_move was set
to true.
- add some color to the test output (minor change)
Previously DC_EVENT_ERROR_NETWORK events were emitted for each failed
attempt during autoconfiguration, even if eventually configuration
succeeded. Such error reports are not useful and often confusing,
especially if they report failures to connect to domains that don't
exist, such as imap.example.org when mail.example.org should be used.
Now DC_EVENT_ERROR_NETWORK is only emitted when attempting to connect
with existing IMAP and SMTP configuration already stored in the
database.
Configuration failure is still indicated by DC_EVENT_CONFIGURE_PROGRESS
with data1 set to 0. Python tests from TestOnlineConfigurationFails
group are changed to only wait for this event.
