Problem: Not easy to start Vim cleanly without changing the viminfo file.
Not possible to know whether the -i command line flag was used.
Solution: Add the --clean command line argument. Add the 'viminfofile'
option. Add "-u DEFAULTS".
Problem: If a job writes to a buffer and the user is typing a command, the
screen isn't updated. When a message is displayed the changed
buffer may cause it to be cleared. (Ramel Eshed)
Solution: Update the screen and then the command line if the screen didn't
scroll. Avoid inserting screen lines, as it clears any message.
Update the status line when the buffer changed.
Problem: job_stop() often assumes the channel will be closed, while the job
may not actually be stopped. (Martin Gammelsæter)
Solution: Only assume the job stops on "kill". Don't send a signal if the
job has already ended. (closes#1632)
Problem: When a json message arrives in pieces, the start is dropped and
the decoding fails.
Solution: Do not drop the start when it is still needed. (Kay Zheng) Add a
test. Reset the timeout when something is received.
Problem: JSON with a duplicate key gives an internal error. (Lcd)
Solution: Give a normal error. Avoid an error when parsing JSON from a
remote client fails.
Problem: When reading channel output in a timer, messages may go missing.
(Skywind)
Solution: Add the "drop" option. Write error messages in the channel log.
Don't have ch_canread() check for the channel being open.
Problem: When using ch_read() with zero timeout, can't tell the difference
between reading an empty line and nothing available.
Solution: Add ch_canread().
Problem: When a channel callback consumes a lot of time Vim becomes
unresponsive. (skywind)
Solution: Bail out of checking channel readahead after 100 msec.
Problem: When the channel callback gets job info the job may already have
been deleted. (lifepillar)
Solution: Do not delete the job when the channel is still useful. (ichizok,
closes#1242, closes#1245)
Problem: An exiting job is detected with a large latency.
Solution: Check for pending job more often. (Ozaki Kiichi) Change the
double loop in mch_inchar() into one.
Problem: A channel is closed when reading on stderr or stdout fails, but
there may still be something to read on another part.
Solution: Turn ch_to_be_closed into a bitfield. (Ozaki Kiichi)
Problem: If a channel in NL mode is missing the NL at the end the remaining
characters are dropped.
Solution: When the channel is closed use the remaining text. (Ozaki Kiichi)
Problem: When using ":sleep" channel input is not handled.
Solution: When there is a channel check for input also when not in raw mode.
Check every 100 msec.
Problem: Can't tell which part of a channel has "buffered" status.
Solution: Add an optional argument to ch_status(). Let ch_info() also
return "buffered" for out_status and err_status.
Problem: The "Reading from channel output..." message can be unwanted.
Appending to a buffer leaves an empty first line behind.
Solution: Add the "out_msg" and "err_msg" options. Writing the first line
overwrites the first, empty line.
Problem: When starting a job that reads from a buffer and reaching the end,
the job hangs.
Solution: Close the pipe or socket when all lines were read.
Problem: Using function() with a name will find another function when it is
redefined.
Solution: Add funcref(). Refer to lambda using a partial. Fix several
reference counting issues.
Problem: Using submatch() in a lambda passed to substitute() is verbose.
Solution: Use a static list and pass it as an optional argument to the
function. Fix memory leak.
Problem: When writing buffer lines to a job there is no NL to NUL
conversion.
Solution: Make it work symmetrical with writing lines from a job into a
buffer.
Problem: When using a job in raw mode to append to a buffer garbage
characters are added.
Solution: Do not replace the trailing NUL with a NL. (Ozaki Kiichi)
Problem: Collapsing channel buffers and searching for NL does not work
properly. (Xavier de Gary, Ramel Eshed)
Solution: Do not assume the buffer contains a NUL or not. Change NUL bytes
to NL to avoid the string is truncated.
Problem: Whether a job has exited isn't detected until a character is
typed. After calling exit_cb the cursor is in the wrong place.
Solution: Don't wait forever for a character to be typed when there is a
pending job. Update the screen if neede after calling exit_cb.
Problem: When a channel writes to a buffer it doesn't find a buffer by the
short name but re-uses it anyway.
Solution: Find buffer also by the short name.
Problem: When a channel appends to a buffer that is 'nomodifiable' there is
an error but appending is done anyway.
Solution: Add the 'modifiable' option. Refuse to write to a 'nomodifiable'
when the value is 1.
