Test_deadly_signal_TERM() fails in FreeBSD CI

17 views
Skip to first unread message

Dominique Pellé

unread,
May 12, 2020, 4:23:28 PM5/12/20
to vim_dev
Hi

I see that Test_deadly_signal_TERM() added in
vim-8.2.0742 fails on FreeBSD CI:
https://cirrus-ci.com/task/6347659038425088

I don't why it's different than on Linux or macOs where
the test passes. Does anybody have a FreeBSD host
to reproduce?

I see that "Caught deadly signal..." is meant to be
output in out_str() in misc1.c. Maybe that output
is not properly flushed?

In test logs:

From test_signals.vim:
Found errors in Test_deadly_signal_TERM():
Run 1:
function RunTheTest[39]..Test_deadly_signal_TERM[20]..WaitForAssert[2]..<SNR>7_WaitForCommon[11]..<lambda>3
line 1: Expected 'Vim: Caught deadly signal TERM' but got '~\[ occurs
74 times]'
function RunTheTest[39]..Test_deadly_signal_TERM[21]..WaitForAssert[2]..<SNR>7_WaitForCommon[11]..<lambda>4
line 1: Pattern 'Vim: preserving files\\.\\.\\.$' does not match '~\[
occurs 74 times]'

Yegappan Lakshmanan

unread,
May 12, 2020, 9:09:02 PM5/12/20
to vim_dev
Hi,

I reproduced this test failure in a FreeBSD setup. The reason the test
is failing is that the messages (Vim: Caught deadly signal TERM, etc.)
are displayed at the bottom of the window on FreeBSD. On MacOS
and other systems, it is displayed at the top of the window.

The term_getline() calls in the test needs to be adjusted for this.

- Yegappan
 

Dominique Pellé

unread,
May 13, 2020, 1:28:18 AM5/13/20
to vim_dev
Thanks Yegappan

If the position where "Vim: Caught deadly signal TERM"
is printed is different, then I think we're better off removing
checking for that message in the test. After sending the
TERM signal, the test verifies anyway that vim terminated
and the swap file was created. That should be enough.

Dominique
Reply all
Reply to author
Forward
0 new messages