Some servers (e.g. Dovecot) use placeholder strings instead of the empty
string (e.g. "MISSING_DOMAIN") in FETCH Envelope responses when a
required address part (e.g. local part, domain) are empty.
This adds a quirk that can be enabled for such servers to check for the
placeholders and if found, replace them with empty strings.
Add a quirks object as a property so code that parses the response has
access to it. Simplify constructing server responses slightly and pass
the IMAP connection's quirks when doing so.
Rename append method to concatenate_list since that's what it actually does. Add new method
for cat'ing a single mailbox, add methods for merging both a single mailbox and mailbox
list into a new list.
If an address does not have an `@`, or if the IMAP constructor is called
with empty mailbox or domain parts, then these properties will be empty.
This is not uncommon, especially on UNIX hosts where system accounts
send email.
Ensure this is handled correctly in the constructors and are
round-tripped correctly by `to_rfc822_address` and hence
`to_rfc822_string`.
Although when processing mailbox addresses lists are effectively sets,
the ordering can be important for people. As such, make `equal_to`
comparisons require identical ordering, and add new `contains_all` for
order-independent comparisons.
Engine mocks don't need to be in the `Geary` namespace, and including
them there makes it difficult to use them in client tests, so put them
all in their own name-space and corresponding directory.
Remove `Message.without_bcc` ctor since we can now get GMime to exclude
BCC headers when serialising, avoiding the overhead of taking a complete
copy of the message just to strip BCCs.
Rename `get_network_buffer` to `get_rfc822_buffer` to be a bit more
explicit in that's the way to get an actual RFC822 formatted message.
Replace book args with flags, and take a protocol-specific approach
rather than feature-specific, because in the end you're either going to
want all the SMTP formatting quirks or none of them.
Remove custom dot-stuffing support from Smtp.ClientConnection, since it
is redundant.
Make constructors use the Message.from_gmime where possible. Merge that
and stock_from_gmime so it's clear the stocking only happens in the
ctor. Update stocking to use high-level GMime APIs where available to
avoid re-parsing header values.
Split Geary.RFC822.MessageData interface up into DecodedMessageData and
EncodedMessageData so the difference between the two is clear and they
can't be used interchangeably.
Add `DecodedMessageData::to_rfc822_string` to provide a common interface
for round-tripping decoded data.
Update all classes to implement one of these and follow the same general
API patterns.
GMail doesn't seem to quote IMAP email flags when they contain spaces
or `]`, a reserved character under RFC 3501. We can't do anything
about the former, but we can work around the latter.
This patch adds a quirk allowing `]` in IMAP flags in general for
GMail accounts, and adds some additional unit tests to that end.
Ensure that if stopping the conversation monitor while it is waiting
for the folder to open that the folder isn't closed, and if an error
is thrown when opening the folder that it gets thrown by
::start_monitoring anyway.
Hopefully fixes#741
Add `create_personal_folder` method, implement it in
ImapEngine.GenericAccount. Use this when creating required special
folders to reduce code duplication.
The `SpecialFolderType` was somewhat mis-named, since the special use
does not (in most cases) confer any special type, rather it's simply
defines what a particular folder is used /for/.
As such, rename the enum to to `Geary.Folder.SpecialUse`, moving it in
to the `Folder` class, since it relates specifically to folders, also
rename `Folder::special_folder_type` and
`Folder::special_folder_type_changed` reflect the above and not
duplicate the type name, and similarly rename
`Account::folders_special_type`.
Update the many call sites.
When attaching body parts to a message from a composed email, don't
re-use the transfer encoding used from the plain part for the HTML part,
since the latter may be different if e.g. the HTML contains no line
breaks and hence is a single long line.
See #771
By not closing/flushing the buffer, ::get_best_charset and
::get_best_encoding were not including the last line of the buffer, and
missing single very long lines when guessing the encoding.
Fixes#771
Convert mailbox and mailbox RW/RO state accessors to properties and
rename to be more explicit about what they mean. Remove mailbox arg
from ::getProtocolState() since it's unused.
