[LTP] [PATCH 1/1] kirk: Update to v3.2

[LTP] [PATCH 1/1] kirk: Update to v3.2

[Petr Vorel]

https://github.com/linux-test-project/kirk/releases/tag/v3.2
* Framework is no longer a plugin by @acerv in #86
* sut: add --optimize-sut option by @acerv in #87
* stabilized tests execution for CI

Signed-off-by: Petr Vorel <pvorel@suse.cz>
---
 tools/kirk/kirk-src | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/kirk/kirk-src b/tools/kirk/kirk-src
index 526f81d43e..2eff923cb5 160000
--- a/tools/kirk/kirk-src
+++ b/tools/kirk/kirk-src
@@ -1 +1 @@
-Subproject commit 526f81d43e711b8a2f1ccc7a89e82b96fe090e96
+Subproject commit 2eff923cb5dd97184c21ae47bd9778eab3f184e8
-- 
2.51.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Andrea Cervesato via ltp]

Hi!

Ah thanks, I guess I need to to update the kirk TODO list for the
release because I always forget to send a follow up patch in LTP.

Will do it now.

Reviewed-by: Andrea Cervesato <andrea.cervesato@suse.com>

-- 
Andrea Cervesato
SUSE QE Automation Engineer Linux
andrea.cervesato@suse.com


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Li Wang via ltp]

Hi Andrea,

A quick question on '--framework' changes in v3.2, since before we
use Kirk with set 'ltp:root' to find the correct LTP installation path.
  `kirk --framework ltp:root=${LTPDIR} --run-suite $RUNTEST -v
--json-report $OUTPUTDIR/$RUNTEST.json`

Aafter removing that --framework, I get an error when trying to use
--env for the installation path.
  `kirk  --evn LTPROOT=${LTPDIR} --run-suite $RUNTEST -v --json-report
$OUTPUTDIR/$RUNTEST.json`

===== Log =====
# ./kirk --env LTPROOT=/root/ltp-install/  --run-suite sched -v
--json-report sched.json
Host information

Hostname:   dell-per430-09.gsslab.pek2.redhat.com
Python:     3.12.12 (main, Jan  6 2026, 00:00:00) [GCC 14.3.1 20250617
(Red Hat 14.3.1-2)]
Directory:  /tmp/kirk.root/tmp1a062qkt

Connecting to SUT: default
Error: LTP folder doesn't exist: /opt/ltp

Disconnecting from SUT: default
Session stopped

==============

But once I put the LTPROOT=${LTPDIR} env separately, it works well:
  `LTPROOT=/root/ltp-install/  ./kirk  --run-suite sched -v
--json-report sched.json`

Did I miss something here, or is it a Kirk issue that should be fixed?


-- 
Regards,
Li Wang


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Andrea Cervesato via ltp]

Hi!

On Thu Jan 15, 2026 at 9:08 AM CET, Li Wang wrote:
> Hi Andrea,
>
> A quick question on '--framework' changes in v3.2, since before we
> use Kirk with set 'ltp:root' to find the correct LTP installation path.
>   `kirk --framework ltp:root=${LTPDIR} --run-suite $RUNTEST -v
> --json-report $OUTPUTDIR/$RUNTEST.json`
>
> Aafter removing that --framework, I get an error when trying to use
> --env for the installation path.
>   `kirk  --evn LTPROOT=${LTPDIR} --run-suite $RUNTEST -v --json-report
> $OUTPUTDIR/$RUNTEST.json`
>
> ===== Log =====
> # ./kirk --env LTPROOT=/root/ltp-install/  --run-suite sched -v
> --json-report sched.json
> Host information
>
> Hostname:   dell-per430-09.gsslab.pek2.redhat.com
> Python:     3.12.12 (main, Jan  6 2026, 00:00:00) [GCC 14.3.1 20250617
> (Red Hat 14.3.1-2)]
> Directory:  /tmp/kirk.root/tmp1a062qkt
>
> Connecting to SUT: default
> Error: LTP folder doesn't exist: /opt/ltp
>
> Disconnecting from SUT: default
> Session stopped
>
> ==============
>

This is strange. The --env feature is taking info from shell and
updating its dictionary if it has been defined.

This is exactly why this ticket has been open. Too much confusion
between shell variables and --env parameter. We can't have 2 different
ways to set internal variables.

https://github.com/linux-test-project/kirk/issues/72

> But once I put the LTPROOT=${LTPDIR} env separately, it works well:
>   `LTPROOT=/root/ltp-install/  ./kirk  --run-suite sched -v
> --json-report sched.json`
>
> Did I miss something here, or is it a Kirk issue that should be fixed?


Please use this method that is more consistent, also for CI.

Thanks,
-- 
Andrea Cervesato
SUSE QE Automation Engineer Linux
andrea.cervesato@suse.com


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Li Wang via ltp]

> > ===== Log =====
> > # ./kirk --env LTPROOT=/root/ltp-install/  --run-suite sched -v
> > --json-report sched.json
> > Host information
> >
> > Hostname:   dell-per430-09.gsslab.pek2.redhat.com
> > Python:     3.12.12 (main, Jan  6 2026, 00:00:00) [GCC 14.3.1 20250617
> > (Red Hat 14.3.1-2)]
> > Directory:  /tmp/kirk.root/tmp1a062qkt
> >
> > Connecting to SUT: default
> > Error: LTP folder doesn't exist: /opt/ltp
> >
> > Disconnecting from SUT: default
> > Session stopped
> >
> > ==============
> >
>
> This is strange. The --env feature is taking info from shell and
> updating its dictionary if it has been defined.
>
> This is exactly why this ticket has been open. Too much confusion
> between shell variables and --env parameter. We can't have 2 different
> ways to set internal variables.
>
> https://github.com/linux-test-project/kirk/issues/72

After playing with the ltp.py for a while and assisted by GPT5,
I drafted a simple patch like below, it works from my test, can
Do you think it's worth opening a PR to Kirk?

--- a/libkirk/ltp.py
+++ b/libkirk/ltp.py
@@ -50,7 +50,7 @@ class LTPFramework(Framework):
         self,
         max_runtime: float = 0.0,
         timeout: float = 30.0,
-        env: dict = {},
+        env: Optional[dict] = None,
     ) -> None:
         """
         :param max_runtime: filter out all tests above this time value
@@ -63,7 +63,8 @@ class LTPFramework(Framework):
         self._logger = logging.getLogger("libkirk.ltp")
         self._cmd_matcher = re.compile(r'(?:"[^"]*"|\'[^\']*\'|\S+)')
         self._max_runtime = max_runtime
-        self._root = os.environ.get("LTPROOT", "/opt/ltp")
+        env = env or {}
+        self._root = env.get("LTPROOT") or os.environ.get("LTPROOT")
or "/opt/ltp"
         self._tc_folder = os.path.join(self._root, "testcases", "bin")

         self._env = {
@@ -79,8 +80,7 @@ class LTPFramework(Framework):
             if timeout:
                 self._env["LTP_TIMEOUT_MUL"] = str((timeout * 0.9) / 300.0)

-        if env:
-            self._env.update(env)
+        self._env.update(env)

     async def _read_path(self, channel: ComChannel) -> Dict[str, str]:
         """



-- 
Regards,
Li Wang


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Petr Vorel]

Hi Li,

> > > ===== Log =====
> > > # ./kirk --env LTPROOT=/root/ltp-install/  --run-suite sched -v
> > > --json-report sched.json
> > > Host information

> > > Hostname:   dell-per430-09.gsslab.pek2.redhat.com
> > > Python:     3.12.12 (main, Jan  6 2026, 00:00:00) [GCC 14.3.1 20250617
> > > (Red Hat 14.3.1-2)]
> > > Directory:  /tmp/kirk.root/tmp1a062qkt

> > > Connecting to SUT: default
> > > Error: LTP folder doesn't exist: /opt/ltp

> > > Disconnecting from SUT: default
> > > Session stopped

> > > ==============


> > This is strange. The --env feature is taking info from shell and
> > updating its dictionary if it has been defined.

> > This is exactly why this ticket has been open. Too much confusion
> > between shell variables and --env parameter. We can't have 2 different
> > ways to set internal variables.

> > https://github.com/linux-test-project/kirk/issues/72

> After playing with the ltp.py for a while and assisted by GPT5,
> I drafted a simple patch like below, it works from my test, can
> Do you think it's worth opening a PR to Kirk?

> --- a/libkirk/ltp.py
> +++ b/libkirk/ltp.py
> @@ -50,7 +50,7 @@ class LTPFramework(Framework):
>          self,
>          max_runtime: float = 0.0,
>          timeout: float = 30.0,
> -        env: dict = {},
> +        env: Optional[dict] = None,
>      ) -> None:
>          """
>          :param max_runtime: filter out all tests above this time value
> @@ -63,7 +63,8 @@ class LTPFramework(Framework):
>          self._logger = logging.getLogger("libkirk.ltp")
>          self._cmd_matcher = re.compile(r'(?:"[^"]*"|\'[^\']*\'|\S+)')
>          self._max_runtime = max_runtime
> -        self._root = os.environ.get("LTPROOT", "/opt/ltp")
> +        env = env or {}
> +        self._root = env.get("LTPROOT") or os.environ.get("LTPROOT")

Do I understand you want to get back --env? Would it be too hard for you to use
shell environment variable? The purpose was to simplify the code (Andrea's
concern, which I don't object because he does all the kirk work).

Also, I'm sorry to merge update without your ack, next time I'll wait for the
testing.

Kind regards,
Petr

> or "/opt/ltp"
>          self._tc_folder = os.path.join(self._root, "testcases", "bin")

>          self._env = {
> @@ -79,8 +80,7 @@ class LTPFramework(Framework):
>              if timeout:
>                  self._env["LTP_TIMEOUT_MUL"] = str((timeout * 0.9) / 300.0)

> -        if env:
> -            self._env.update(env)
> +        self._env.update(env)

>      async def _read_path(self, channel: ComChannel) -> Dict[str, str]:
>          """

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Li Wang via ltp]

On Thu, Jan 15, 2026 at 6:04 PM Petr Vorel <pvorel@suse.cz> wrote:
>
> Hi Li,
>
> > > > ===== Log =====
> > > > # ./kirk --env LTPROOT=/root/ltp-install/  --run-suite sched -v
> > > > --json-report sched.json
> > > > Host information
>
> > > > Hostname:   dell-per430-09.gsslab.pek2.redhat.com
> > > > Python:     3.12.12 (main, Jan  6 2026, 00:00:00) [GCC 14.3.1 20250617
> > > > (Red Hat 14.3.1-2)]
> > > > Directory:  /tmp/kirk.root/tmp1a062qkt
>
> > > > Connecting to SUT: default
> > > > Error: LTP folder doesn't exist: /opt/ltp
>
> > > > Disconnecting from SUT: default
> > > > Session stopped
>
> > > > ==============
>
>
> > > This is strange. The --env feature is taking info from shell and
> > > updating its dictionary if it has been defined.
>
> > > This is exactly why this ticket has been open. Too much confusion
> > > between shell variables and --env parameter. We can't have 2 different
> > > ways to set internal variables.
>
> > > https://github.com/linux-test-project/kirk/issues/72
>
> > After playing with the ltp.py for a while and assisted by GPT5,
> > I drafted a simple patch like below, it works from my test, can
> > Do you think it's worth opening a PR to Kirk?
>
> > --- a/libkirk/ltp.py
> > +++ b/libkirk/ltp.py
> > @@ -50,7 +50,7 @@ class LTPFramework(Framework):
> >          self,
> >          max_runtime: float = 0.0,
> >          timeout: float = 30.0,
> > -        env: dict = {},
> > +        env: Optional[dict] = None,
> >      ) -> None:
> >          """
> >          :param max_runtime: filter out all tests above this time value
> > @@ -63,7 +63,8 @@ class LTPFramework(Framework):
> >          self._logger = logging.getLogger("libkirk.ltp")
> >          self._cmd_matcher = re.compile(r'(?:"[^"]*"|\'[^\']*\'|\S+)')
> >          self._max_runtime = max_runtime
> > -        self._root = os.environ.get("LTPROOT", "/opt/ltp")
> > +        env = env or {}
> > +        self._root = env.get("LTPROOT") or os.environ.get("LTPROOT")
>
> Do I understand you want to get back --env? Would it be too hard for you to use
> shell environment variable? The purpose was to simplify the code (Andrea's
> concern, which I don't object because he does all the kirk work).

No, I don't want to get back --env, since before Andrea replied to me
I've beendebugging a while and found a way to make --env work.

So just let him know it's not too complicated to support both way,
I would leave to him to make the final decision :).

> Also, I'm sorry to merge update without your ack, next time I'll wait for the
> testing.

No problem, we're still using the stable LTP version, so today I'm just testing
Kirk-3.2 to see if an upgrade is possible.

-- 
Regards,
Li Wang


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Andrea Cervesato via ltp]

>
> After playing with the ltp.py for a while and assisted by GPT5,
> I drafted a simple patch like below, it works from my test, can
> Do you think it's worth opening a PR to Kirk?

Thanks for checking, see considerations below.

>
> --- a/libkirk/ltp.py
> +++ b/libkirk/ltp.py
> @@ -50,7 +50,7 @@ class LTPFramework(Framework):
>          self,
>          max_runtime: float = 0.0,
>          timeout: float = 30.0,
> -        env: dict = {},
> +        env: Optional[dict] = None,

No need for this, if default it's {} already. That None adds one more
layer of complexity, because then we need to make sure env is not None.

>      ) -> None:
>          """
>          :param max_runtime: filter out all tests above this time value
> @@ -63,7 +63,8 @@ class LTPFramework(Framework):
>          self._logger = logging.getLogger("libkirk.ltp")
>          self._cmd_matcher = re.compile(r'(?:"[^"]*"|\'[^\']*\'|\S+)')
>          self._max_runtime = max_runtime
> -        self._root = os.environ.get("LTPROOT", "/opt/ltp")
> +        env = env or {}

See above.

> +        self._root = env.get("LTPROOT") or os.environ.get("LTPROOT")
> or "/opt/ltp"

This makes sense, but that "or /opt/ltp" is not needed. See below.

>          self._tc_folder = os.path.join(self._root, "testcases", "bin")
>
>          self._env = {
> @@ -79,8 +80,7 @@ class LTPFramework(Framework):
>              if timeout:
>                  self._env["LTP_TIMEOUT_MUL"] = str((timeout * 0.9) / 300.0)
>
> -        if env:
> -            self._env.update(env)
> +        self._env.update(env)
>
>      async def _read_path(self, channel: ComChannel) -> Dict[str, str]:
>          """

diff --git a/libkirk/ltp.py b/libkirk/ltp.py
index 2c0f6d2..153bc7b 100644
--- a/libkirk/ltp.py
+++ b/libkirk/ltp.py
@@ -63,7 +63,7 @@ class LTPFramework(Framework):
         self._logger = logging.getLogger("libkirk.ltp")
         self._cmd_matcher = re.compile(r'(?:"[^"]*"|\'[^\']*\'|\S+)')
         self._max_runtime = max_runtime
-        self._root = os.environ.get("LTPROOT", "/opt/ltp")
+        self._root = env.get("LTPROOT") or os.environ.get("LTPROOT", "/opt/ltp")
         self._tc_folder = os.path.join(self._root, "testcases", "bin")
 
         self._env = {
@@ -94,8 +94,7 @@ class LTPFramework(Framework):
             if not ret or ret["returncode"] != 0:
                 raise FrameworkError("Can't read PATH variable")
 
-            tcases = os.path.join(self._root, "testcases", "bin")
-            env["PATH"] = ret["stdout"].strip() + f":{tcases}"
+            env["PATH"] = ret["stdout"].strip() + f":{self._tc_folder}"
 
         self._logger.debug("PATH=%s", env["PATH"])
 

-- 
Andrea Cervesato
SUSE QE Automation Engineer Linux
andrea.cervesato@suse.com


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Li Wang via ltp]

Andrea Cervesato <andrea.cervesato@suse.com> wrote:

> > After playing with the ltp.py for a while and assisted by GPT5,
> > I drafted a simple patch like below, it works from my test, can
> > Do you think it's worth opening a PR to Kirk?
>
> Thanks for checking, see considerations below.
>
> >
> > --- a/libkirk/ltp.py
> > +++ b/libkirk/ltp.py
> > @@ -50,7 +50,7 @@ class LTPFramework(Framework):
> >          self,
> >          max_runtime: float = 0.0,
> >          timeout: float = 30.0,
> > -        env: dict = {},
> > +        env: Optional[dict] = None,
>
> No need for this, if default it's {} already. That None adds one more
> layer of complexity, because then we need to make sure env is not None.
>
> >      ) -> None:
> >          """
> >          :param max_runtime: filter out all tests above this time value
> > @@ -63,7 +63,8 @@ class LTPFramework(Framework):
> >          self._logger = logging.getLogger("libkirk.ltp")
> >          self._cmd_matcher = re.compile(r'(?:"[^"]*"|\'[^\']*\'|\S+)')
> >          self._max_runtime = max_runtime
> > -        self._root = os.environ.get("LTPROOT", "/opt/ltp")
> > +        env = env or {}
>
> See above.
>
> > +        self._root = env.get("LTPROOT") or os.environ.get("LTPROOT")
> > or "/opt/ltp"
>
> This makes sense, but that "or /opt/ltp" is not needed. See below.
>
> >          self._tc_folder = os.path.join(self._root, "testcases", "bin")
> >
> >          self._env = {
> > @@ -79,8 +80,7 @@ class LTPFramework(Framework):
> >              if timeout:
> >                  self._env["LTP_TIMEOUT_MUL"] = str((timeout * 0.9) / 300.0)
> >
> > -        if env:
> > -            self._env.update(env)
> > +        self._env.update(env)
> >
> >      async def _read_path(self, channel: ComChannel) -> Dict[str, str]:
> >          """
>
> diff --git a/libkirk/ltp.py b/libkirk/ltp.py
> index 2c0f6d2..153bc7b 100644
> --- a/libkirk/ltp.py
> +++ b/libkirk/ltp.py
> @@ -63,7 +63,7 @@ class LTPFramework(Framework):
>          self._logger = logging.getLogger("libkirk.ltp")
>          self._cmd_matcher = re.compile(r'(?:"[^"]*"|\'[^\']*\'|\S+)')
>          self._max_runtime = max_runtime
> -        self._root = os.environ.get("LTPROOT", "/opt/ltp")
> +        self._root = env.get("LTPROOT") or os.environ.get("LTPROOT", "/opt/ltp")
>          self._tc_folder = os.path.join(self._root, "testcases", "bin")
>
>          self._env = {
> @@ -94,8 +94,7 @@ class LTPFramework(Framework):
>              if not ret or ret["returncode"] != 0:
>                  raise FrameworkError("Can't read PATH variable")
>
> -            tcases = os.path.join(self._root, "testcases", "bin")
> -            env["PATH"] = ret["stdout"].strip() + f":{tcases}"
> +            env["PATH"] = ret["stdout"].strip() + f":{self._tc_folder}"
>
>          self._logger.debug("PATH=%s", env["PATH"])


This change worked for me, and I saw in kirk/issues/72 that someone
still wants support for both methods.

So, maybe keeping it can be helpful to set env for remote SUT
execution? Since exporting locally may not affect the remote side
unless Kirk forwards env.

I am not quite familar to Kirk code, so just raise my questions to you
to evaluation.

-- 
Regards,
Li Wang


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Andrea Cervesato via ltp]

>
> This change worked for me, and I saw in kirk/issues/72 that someone
> still wants support for both methods.
>
> So, maybe keeping it can be helpful to set env for remote SUT
> execution? Since exporting locally may not affect the remote side
> unless Kirk forwards env.
>
> I am not quite familar to Kirk code, so just raise my questions to you
> to evaluation.

The problem to keep both is the reason why these bugs are happening, so I
would just drop --env. An alternative is to drop variables completely
and to use the same method of runltp, where -d and -r are used to set
TMPDIR and LTPROOT.

The reason why --env was added is that we might wanted to have multiple
frameworks, but now that we support LTP only there's no this issue
anymore.

-- 
Andrea Cervesato
SUSE QE Automation Engineer Linux
andrea.cervesato@suse.com


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Andrea Cervesato via ltp]

I updated kirk master branch, if there are no more requests, I could
create a v3.2.1 during the pre-release process of LTP.

I will wait and see if there are more bugs to fix and eventually create
a new release.


-- 
Andrea Cervesato
SUSE QE Automation Engineer Linux
andrea.cervesato@suse.com


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] kirk: Update to v3.2

[Li Wang via ltp]

Awesome, thanks for the update!

On Thu, Jan 15, 2026 at 9:02 PM Andrea Cervesato
<andrea.cervesato@suse.com> wrote:
>
> I updated kirk master branch, if there are no more requests, I could
> create a v3.2.1 during the pre-release process of LTP.
>
> I will wait and see if there are more bugs to fix and eventually create
> a new release.
>
>
> --
> Andrea Cervesato
> SUSE QE Automation Engineer Linux
> andrea.cervesato@suse.com
>


-- 
Regards,
Li Wang


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp