Diff Detail
- Repository
- rG FreeBSD src repository
- Lint
Lint Skipped - Unit
Tests Skipped - Build Status
Buildable 67198 Build 64081: arc lint + arc unit
Event Timeline
The prototype has the following logic currently:
1 Invoking w/o argument runs $SHELL (falls back to hardcoded /usr/bin/ if SHELL env var is not found). It does chdir to the test's working dir:
> kyua debug -x test1 > kyua debug --execute test1 root@taarch64:/usr/tests/sys/netpfil/pf # kyua debug -x mbuf:inet_in_mbuf_len root@taarch64:/tmp/kyua.h6W16I/2/work # ls -a . .. created_interfaces.lst created_jails.lst root@taarch64:/tmp/kyua.h6W16I/2/work # exit Executing command [ ping -c1 192.0.2.2 ] alcatraz: removed Fail: incorrect exit status: 0, expected: 1 stdout: PING 192.0.2.2 (192.0.2.2): 56 data bytes 64 bytes from 192.0.2.2: icmp_seq=0 ttl=64 time=0.097 ms --- 192.0.2.2 ping statistics --- 1 packets transmitted, 1 packets received, 0.0% packet loss round-trip min/avg/max/stddev = 0.097/0.097/0.097/0.000 ms stderr: Files left in work directory after failure: created_interfaces.lst, created_jails.lst ifconfig: interface epair0a does not exist mbuf:inet_in_mbuf_len -> failed: atf-check failed; see the output of the test for details root@taarch64:/usr/tests/sys/netpfil/pf # echo $? 1
2 If an argument is passed then $SHELL (falls back to hard-coded /bin/sh) is run with the given command string (aka $SHELL -c $cmd). A shell is expected to have -c option. It also does chdir to the test's working dir:
> kyua debug -x'printf "\a"' test1 > kyua debug --execute="cat some_file_from_test_working_dir; do_something_else" test1
The default string is intentionally set to $SHELL literal to be a hint during help output reading:
root@taarch64:/usr/tests/sys/netpfil/pf # kyua help debug
kyua (Kyua) 0.13
Usage: kyua [general_options] debug [command_options] test_case
...
Available command options:
--build-root=path Path to the built test programs, if different from the location of the
Kyuafile scripts.
-k file, --kyuafile=file Path to the test suite definition (default: Kyuafile).
-p, --pause-before-cleanup-upon-fail Pauses right before the test cleanup upon fail.
--pause-before-cleanup Pauses right before the test cleanup.
--stdout=path Where to direct the standard output of the test case (default: /dev/stdout).
--stderr=path Where to direct the standard error of the test case (default: /dev/stderr).
-x cmd, --execute=cmd A command to run upon test failure (default: $SHELL).
See kyua-debug(1) for more details.P.S. Eventually, I plan to extract base_option/string_option arg optionality to separate commit(s).
This works almost exactly how I want, thank you. I have one "complaint": when I use this mode, kyua 1) drops me in the shell, 2) waits for the shell to exit, 3) prints the debug output, e.g., executed commands and the test failure string. The -p mode is basically the same. It is more useful to do 3) first, then 1) and 2). Is it hard to make this change?
Yes, both kyua debug -x and kyua debug -p have the same behavior as the same internal mechanism is used. The following patch addresses this "inconvenience by design": https://reviews.freebsd.org/D54363
| contrib/kyua/cli/cmd_debug.cpp | ||
|---|---|---|
| 105 | The next version of kyua should focus on more modern standards (C++-17 at least; ideally C++-20). With that in mind you can technically use std::filesystem_path::current_path(..) instead of chdir(2). I'm trying to wrangle everything together upstream since I have some cycles right now. | |
This works well, thanks. It would probably be useful to document/log the fact that if the test uses a execenv jail, then -x will attach the command to the jail.
| contrib/kyua/cli/cmd_debug.cpp | ||
|---|---|---|
| 105 | What's wrong with just using chdir()? Is kyua going to run on windows? | |