Subscription status was checked/modified in various places, whereas now
the universal truth is the `BackendFolder::is_subscribed()` method set
by the backend when a folder is created. The `Account` struct passes a
closure to the backend constructor that determines whether the folder is subscribed or not according to the user configuration.
- If subscribed_folders field is empty, then all folders are subscribed.
- OR check explicit folder configuration
- OR check if folder path matches to a glob in subscribed_folders.
`DisplayMessage` messages are for user input responses (eg errors for
user actions). They now appear as floating boxes in the bottom right
corner of the UI and can be browsed with Alt('<') and Alt('>')
No logical reason for it not to be in the terminal module anymore (the
set_and_join* functions predate the terminal module which is why they
weren't there to begin with).
Merge ui crate with root crate.
In preparation for uploading `meli` as a separate crate on crates.io.
Workspace crates will need to be published as well and having a separate
`ui` crate and binary perhaps doesn't make sense anymore.
First line of an mbox message is a "From ..." header without the colon
required in RFC822. Skip it when parsing the actual e-mail.
This was lost somewhere in the commit history when mbox was added,
weird.
- Add BINDIR, MANDIR env vars
- add check-dep target that checks for cargo/rustc version
- add ANSI formatting output with NO_COLOR support
- add help target
- move manpage installation to install-doc target
- move bin installation to install-bin target
- add dist target
- add warning if BINDIR is not inside $PATH
- add warning if MANDIR is not inside $MANPATH/manpath
This seems to be a bug with debian's troff renderer; tables spanning
more than one page were silently dropped and were not visible.
tbl(1) says to use the macro .TS H with .TH to define the headers but
this is not supported in debian nor openbsd's mandoc implementation.
`Attr` (terminal attributes such as bold, underline, etc) were not being
printed properly: their bitmap representation was printed instead of the
correct ANSI codes to turn them on/off. This worked so far because the
attributes and {fore,back}ground color was reset in every character
print.
draw_horizontal_segment() now keeps state of current_{fg,bg,attr} to
keep from resetting in each column draw.
When inserting reply, its thread group was re-inserted with the reply as
the root. This is a mistake as threads should never be re-inserted, only
modified.