Problem:
The "restart" event has some problems:
- all UI clients must implement a somewhat complex set of setups
- UI must be on the same machine as the server
- only works for the "current" UI
- race/edge case: If the user config has errors / waiting for input, are
all UIs able to attach while Nvim is waiting for input?
Solution:
- Perform the restart on the server, not the client.
- Pass listen address (instead of CLI args) in the UI event.
- Simplifies UI logic: they only need to attach to new address.
- Opens the door for more enhancements in the future, such as allowing
all UIs to reattach instead of only the "current" UI.
Co-authored-by: zeertzjq <zeertzjq@outlook.com>
Co-authored-by: Justin M. Keyes <justinkz@gmail.com>
Problem:
When stopping a PTY process on Windows, the exit code indicates that the
process is stopped by SIGTERM even when closing all streams is enough to
terminate the process. This is inconsistent with other platforms.
Solution:
Set exit_signal to SIGHUP instead of SIGTERM when using SIGHUP.
This reverts commit ab8371a26c.
Need to think of a different solution, which may require adding new
flags to nvim_ui_attach() (e.g. passing stdout or stderr fd).
Problem: If Nvim server fails to --listen and prints error before the
TUI enters alternate screen, the error isn't visible.
Solution: Forward server stderr using client side stderr handler instead
of having the server inherit client stderr file descriptor.
This does mean that `stderr_isatty` will be `false` in the server, but
that value doesn't matter in embedded mode.
Always pass `stdin_fd` to embedded server to avoid a hang when reading
from stdin when it's a TTY (not sure why one wants to do that, perhaps
by mistake), because if `stdin_fd` isn't passed, the server will try to
use stderr as stdin.
Example test failure on CI:
FAILED test/functional/terminal/tui_spec.lua @ 41: TUI exit status 1 and error message with server --listen error #34365
test/functional/terminal\tui_spec.lua:55: Failed to match any screen lines.
Expected (anywhere): "nvim%.exe: Failed to %-%-listen: address already in use:"
Actual:
|{114:nvim.exe -h"} |
| |
|[Process exited 1]^ |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
|{5:-- TERMINAL --} |
Snapshot:
screen:expect([[
{114:nvim.exe -h"} |
|
[Process exited 1]^ |
|*13
{5:-- TERMINAL --} |
]])
stack traceback:
test\functional\ui\screen.lua:909: in function '_wait'
test\functional\ui\screen.lua:537: in function 'expect'
test/functional/terminal\tui_spec.lua:55: in function <test/functional/terminal\tui_spec.lua:41>
In this case, it appears that the client entered alternate screen in the
middle of the server's print_mainerr().
It is indeed possible that a signal may arrive during flush_stream()
(e.g. SIGCHLD from another child process), so poll() again on EINTR.
Fixes the following Coverity warning:
_____________________________________________________________________________________________
*** CID 644345: Error handling issues (CHECKED_RETURN)
/src/nvim/event/proc.c: 405 in flush_stream()
399
400 #ifdef __linux__
401 // On Linux, libuv's polling (which uses epoll) doesn't flush PTY master's pending
402 // work on kernel workqueue, so use an explcit poll() before that. #37982
403 if (proc->type == kProcTypePty && !stream->did_eof) {
404 struct pollfd pollfd = { .fd = ((PtyProc *)proc)->tty_fd, .events = POLLIN };
>>> CID 644345: Error handling issues (CHECKED_RETURN)
>>> Calling "poll(&pollfd, 1UL, 0)" without checking return value. This library function may fail and return an error code.
405 poll(&pollfd, 1, 0);
406 }
407 #endif
408 // Poll for data and process the generated events.
409 loop_poll_events(proc->loop, 0);
410 if (stream->s.events) {
Another possible error is ENOMEM, which is probably not worth handling.
For reference, #23308 previously removed a case of ENOMEM handling, and
this PR also removes an outdated mention of that.
Also reduce the number of #ifdefs in non-OS-specific files.
Using uv_pipe_t to read from PTY master may drop data if the PTY process
exits immediately after output, as libuv treats a partial read after
POLLHUP as EOF, which isn't true for PTY master on Linux. Therefore use
uv_poll_t instead.
This commit can be reverted if libuv/libuv#4992 is fixed for uv_pipe_t.
These are not needed after #35129 but making uncrustify still play nice
with them was a bit tricky.
Unfortunately `uncrustify --update-config-with-doc` breaks strings
with backslashes. This issue has been reported upstream,
and in the meanwhile auto-update on every single run has been disabled.
Uses the undocumented "error_exit" UI event for a different purpose:
When :detach is used on the server, send an "error_exit" with 0 `status`
to indicate that the server shouldn't wait for client exit.
Problem:
On Windows, since b360c06085, UI may hang on exit. #33019
Solution:
Restore the hack in on_proc_exit, until we can figure out why rpc_close
is not called in the UI client when the server closes the channel.
Problem:
on_proc_exit() has a special-case that assumes that the UI client will
never spawn more than 1 child process.
Solution:
If the Nvim server exits, the stream EOF will trigger `rpc_close()` in
the UI client, so we don't need the special case in `on_proc_exit`.
Pass `Channel.exit_status` from `rpc_close()` so that the correct exit
code is reflected.
Problem:
On Windows, spawning the `nvim --embed` server with `detach=true` breaks
various `tt.setup_child_nvim` tests.
Solution:
Make this behavior opt-in with an env var, temporarily.
loop->children might have been a linked list because used to be
modified in place while looped over. However the loops that exists
rather schedules events to be processed later, outside of the loop,
so this can not happen anymore.
When a linked list is otherwise useful it is better to use
lib/queue_defs.h which defines an _intrusive_ linked list (i e it
doesn't need to do allocations for list items like klist ).
Problem:
- "process" is often used as a verb (`multiqueue_process_events`), which
is ambiguous for cases where it's used as a topic.
- The documented naming convention for processes is "proc".
- `:help dev-name-common`
- Shorter is better, when it doesn't harm readability or
discoverability.
Solution:
Rename "process" => "proc" in all C symbols and module names.
