It is perfectly fine and expected to detach from the screen just by
the UI disconnecting from nvim or exiting nvim. Just keep detach() in
screen_basic_spec, to get some coverage of the detach method itself.
This avoids hang on failure in many situations (though one could argue
that detach() should be "fast", or at least "as fast as resize",
which works in press-return already).
Never use detach() just to change the size of the screen, try_resize()
method exists for that specifically.
Problem: If a job exits while waiting on another job, the on_exit
handler is queued but f_jobwait() skips it.
Solution: Always do process_wait(), so that handlers are run during
f_jobwait().
fix#8302
Test case:
$ BUSTED_ARGS="--repeat=2000 --no-keep-going" TEST_FILE=test/functional/core/job_spec.lua TEST_FILTER=waiting make functionaltest
Failure example (macOS CI):
FAILED test/functional/core/job_spec.lua: jobs jobwait will run callbacks while waiting
test/functional/core/job_spec.lua:606: Expected objects to be the same.
Passed in:
(table: 0x1be77c80) {
[1] = 'notification'
[2] = 'wait'
*[3] = {
*[1] = 3 } }
Expected:
(table: 0x1be77d10) {
[1] = 'notification'
[2] = 'wait'
*[3] = {
*[1] = 4 } }
stack traceback:
test/functional/core/job_spec.lua:606: in function <test/functional/core/job_spec.lua:583
Test sometimes fails on macOS CI:
FAILED test/functional/core/job_spec.lua: jobs jobwait will run callbacks while waiting
test/functional/core/job_spec.lua:606: Expected objects to be the same.
Passed in:
(table: 0x1be77c80) {
[1] = 'notification'
[2] = 'wait'
*[3] = {
*[1] = 3 } }
Expected:
(table: 0x1be77d10) {
[1] = 'notification'
[2] = 'wait'
*[3] = {
*[1] = 4 } }
stack traceback:
test/functional/core/job_spec.lua:606: in function <test/functional/core/job_spec.lua:583
Change the test to check if all jobs are starting, not only exiting.
Adapt some tests for OpenBSD:
- scrollback_spec:
- seq(1) is not available on OpenBSD: we'd use jot(1).
- Instead use a (hopefully) portable awk(1) snippet.
- channels_spec
- job_spec
- tui_spec
There was a longer timeout for Windows already, but unlike stated in
51d42917f it is not a worst-case, but gets waited for always.
The test is only about "-1" on timeout, so reduce it to this.
Fixes:
16:33:19,309 INFO - not ok 627 - jobs jobwait with timeout argument will return -1 if the wait timed out
16:33:19,309 INFO - # test/functional/core/job_spec.lua @ 707
16:33:19,309 INFO - # Failure message: test/functional/core/job_spec.lua:714: Expected objects to be the same.
16:33:19,309 INFO - # Passed in:
16:33:19,309 INFO - # (table: 0x0db1a3f0) {
16:33:19,309 INFO - # [1] = 'notification'
16:33:19,309 INFO - # [2] = 'wait'
16:33:19,309 INFO - # *[3] = {
16:33:19,309 INFO - # *[1] = {
16:33:19,309 INFO - # *[1] = -1
16:33:19,309 INFO - # [2] = -1 } } }
16:33:19,309 INFO - # Expected:
16:33:19,309 INFO - # (table: 0x0db1a480) {
16:33:19,309 INFO - # [1] = 'notification'
16:33:19,309 INFO - # [2] = 'wait'
16:33:19,309 INFO - # *[3] = {
16:33:19,309 INFO - # *[1] = {
16:33:19,309 INFO - # *[1] = 4
16:33:19,309 INFO - # [2] = -1 } } }
16:33:19,309 INFO - # stack traceback:
16:33:19,309 INFO - # test/functional/core/job_spec.lua:714: in function <test/functional/core/job_spec.lua:707>
* os/fs.c: add os_isdir_executable()
* eval.c: fix hang on job start caused by non-executable cwd option
* channel.c: assert cwd is an executable directory
* test: jobstart() produces error when using non-executable cwd
likely fixes#7768#7913
If multiple internal stream callbacks were recieved before vimL
callbacks got called, only invoke one vimL callback with all data.
cmd.exe (shell) is faster and more reliable than powershell (.NET frontend).
It's best for short and basic tests that don't require non-trivial scripting.
cmd.exe doesn't support sleep so use powershell's Start-Sleep as substitute.
- echo "" does not hang in powershell
- cmd.exe's echo command does not hang.
- job tests default to powershell (WHY?)
- wait 5 seconds for powershell to create an empty file
- powershell is slow
- cannot reliably validate the id returned by jobstart via jobpid, jobstop
- if using cmd.exe, waiting for a second should be enough
- remaining job tests are unreliable in Windows because any build can pass/fail
for same conditions without changes, especially if the error is in stderr
Hope this will make people using feed_command less likely: this hides bugs.
Already found at least two:
1. msgpackparse() will show internal error: hash_add() in case of duplicate
keys, though it will still work correctly. Currently silenced.
2. ttimeoutlen was spelled incorrectly, resulting in option not being set when
expected. Test was still functioning somehow though. Currently fixed.
