On 12.07.2012 [21:47:11 -0300], Cleber Rosa wrote:
> On 07/12/2012 08:28 PM, Cleber Rosa wrote:
> >On 07/12/2012 08:14 PM, Nishanth Aravamudan wrote:
> >>On 12.07.2012 [20:05:14 -0300], Cleber Rosa wrote:
> >>>On 07/12/2012 04:46 PM, Nishanth Aravamudan wrote:
> >>>>On 12.07.2012 [13:33:16 -0300], Cleber Rosa wrote:
> >>>>>On 07/12/2012 08:55 AM, Cleber Rosa wrote:
> >>>>>>On 07/12/2012 05:07 AM, Nishanth Aravamudan wrote:
> >>>>>>>On 12.07.2012 [00:20:04 -0700], Nishanth Aravamudan wrote:
> >>>>>>>>Hi all,
> >>>>>>>>
> >>>>>>>>After finally bothering to just put a proxy in place locally, I got
> >>>>>>>>grubby 8.15 to build, but it doesn't seem to make boottool happy :)
> >>>>>>>>
> >>>>>>>>07/12 02:53:12 INFO | boottool:0667| Installing grubby
> >>>>>>>>because currently installed version (8.11) is not recent
> >>>>>>>>enough
> >>>>>>>>07/12 02:53:18 DEBUG| boottool:0674| Installed: grubby version 8.15
> >>>>>>>>07/12 02:53:18 ERROR| job:1338| JOB ERROR: Unhandled
> >>>>>>>>TypeError: sequence item 0: expected string, NoneType found
> >>>>>>>>Traceback (most recent call last):
> >>>>>>>> File "/usr/local/autotest/job.py", line 1089, in _run_step_fn
> >>>>>>>> exec('__ret = %s(*__args, **__dargs)' % fn, local_vars,
> >>>>>>>>local_vars)
> >>>>>>>> File "<string>", line 1, in <module>
> >>>>>>>> File "/usr/local/autotest/control.autoserv", line 16, in
> >>>>>>>> boot_kernel
> >>>>>>>> host.cleanup_kernels()
> >>>>>>>> File "/usr/local/autotest/shared/hosts/base_classes.py",
> >>>>>>>>line 655, in cleanup_kernels
> >>>>>>>> boot_info = self.bootloader.get_entries()
> >>>>>>>> File "/usr/local/autotest/tools/boottool.py", line 1150, in
> >>>>>>>>get_entries
> >>>>>>>> raw = self.get_info()
> >>>>>>>> File "/usr/local/autotest/tools/boottool.py", line 1180, in
> >>>>>>>> get_info
> >>>>>>>> info = self._run_grubby_get_output([command])
> >>>>>>>> File "/usr/local/autotest/tools/boottool.py", line 835, in
> >>>>>>>>_run_grubby_get_output
> >>>>>>>> return self._run_get_output(args)
> >>>>>>>> File "/usr/local/autotest/tools/boottool.py", line 751, in
> >>>>>>>>_run_get_output
> >>>>>>>> ' '.join(arguments))
> >>>>>>>>TypeError: sequence item 0: expected string, NoneType found
> >>>>>>>>
> >>>>>>>>07/12 02:53:18 INFO | job:0210| END ABORT ---- ----
> >>>>>>>>timestamp=1342075998 localtime=Jul 12 02:53:18 Unhandled
> >>>>>>>>TypeError: sequence item 0: expected string, NoneType found
> >>>>>>>> Traceback (most recent call last):
> >>>>>>>> File "/usr/local/autotest/job.py", line 1089, in _run_step_fn
> >>>>>>>> exec('__ret = %s(*__args, **__dargs)' % fn, local_vars,
> >>>>>>>>local_vars)
> >>>>>>>> File "<string>", line 1, in <module>
> >>>>>>>> File "/usr/local/autotest/control.autoserv", line 16, in
> >>>>>>>>boot_kernel
> >>>>>>>> host.cleanup_kernels()
> >>>>>>>> File "/usr/local/autotest/shared/hosts/base_classes.py",
> >>>>>>>>line 655, in cleanup_kernels
> >>>>>>>> boot_info = self.bootloader.get_entries()
> >>>>>>>> File "/usr/local/autotest/tools/boottool.py", line 1150,
> >>>>>>>>in get_entries
> >>>>>>>> raw = self.get_info()
> >>>>>>>> File "/usr/local/autotest/tools/boottool.py", line 1180,
> >>>>>>>>in get_info
> >>>>>>>> info = self._run_grubby_get_output([command])
> >>>>>>>> File "/usr/local/autotest/tools/boottool.py", line 835,
> >>>>>>>>in _run_grubby_get_output
> >>>>>>>> return self._run_get_output(args)
> >>>>>>>> File "/usr/local/autotest/tools/boottool.py", line 751,
> >>>>>>>>in _run_get_output
> >>>>>>>> ' '.join(arguments))
> >>>>>>>> TypeError: sequence item 0: expected string, NoneType found
> >>>>>>>> 07/12 02:53:18 DEBUG| base_job:0349| Persistent state
> >>>>>>>>client._record_indent now set to 0
> >>>>>>>I won't claim to understand why it's happening, but I can tell a bit
> >>>>>>>more about what's happening:
> >>>>>>>
> >>>>>>>Adding some hacks locally to _run_grubby_get_output, to print out the
> >>>>>>>parameter arguments and the prepared return value args, I get:
> >>>>>>>
> >>>>>>>04:01:35 INFO | Writing results to /usr/local/autotest/results/default
> >>>>>>>04:01:35 INFO | START ---- ---- timestamp=1342080095
> >>>>>>>localtime=Jul 12 04:01:35
> >>>>>>>04:01:36 INFO | arguments : ['--info=ALL']
> >>>>>>>04:01:36 INFO | self.path = None
> >>>>>>>04:01:36 INFO | arguments : ['--version']
> >>>>>>>04:01:36 INFO | self.path = /sbin/grubby
> >>>>>>>04:01:36 INFO | args : ['/sbin/grubby', '--version']
> >>>>>>>04:01:36 INFO | arguments : ['--version']
> >>>>>>>04:01:36 INFO | self.path = /sbin/grubby
> >>>>>>>04:01:36 INFO | args : ['/sbin/grubby', '--version']
> >>>>>>>04:01:36 INFO | arguments : ['--version']
> >>>>>>>04:01:36 INFO | self.path = /sbin/grubby
> >>>>>>>04:01:36 INFO | args : ['/sbin/grubby', '--version']
> >>>>>>>04:01:36 INFO | args : [None, '--grub2', '--info=ALL']
> >>>>>>>04:01:36 ERROR| JOB ERROR: Unhandled TypeError: sequence item 0:
> >>>>>>>expected string, NoneType found
> >>>>>>>
> >>>>>>>So a couple of questions come to mind:
> >>>>>>>
> >>>>>>>1) why do we call --version so many times?
> >>>>>>There's certainly room for optimizations here.
> >>>>>>
> >>>>>>>2) AFAICT from my debugging, this is all coming from the one call to
> >>>>>>>get_info(), but I don't see how self.path can be None.
> >>>>>>I've seen the same error when I had two grub entries with the same
> >>>>>>kernel and parameters set. I'm *guessiing* this is related to it,
> >>>>>>as your configuration also seems to be like that. I'm debugging
> >>>>>>that right now and will let you know about it.
> >>>>>Nish,
> >>>>>
> >>>>>My first guess was actually incorrect. I've identified that the
> >>>>>error is related to the laziness of the client/shared/boottool.py
> >>>>>code when instantiating boottool, which inherits from Grubby.
> >>>>>
> >>>>>I've switched from the lazy approach to simply importing Grubby as
> >>>>>boottool from the boottool_cli module:
> >>>>>
> >>>>>from boottool_cli import Grubby as boottool
> >>>>>and the error ceases to exist.
> >>>>>
> >>>>>I'm working on fixing that now.
> >>>>Something like the following?
> >>>>
> >>>>diff --git a/client/shared/boottool.py b/client/shared/boottool.py
> >>>>index 90f936b..05858d6 100644
> >>>>--- a/client/shared/boottool.py
> >>>>+++ b/client/shared/boottool.py
> >>>>@@ -29,30 +29,4 @@ except ImportError:
> >>>> imp.load_source("boottool_cli", BOOTTOOL_CLI_PATH)
> >>>>-from boottool_cli import Grubby, install_grubby_if_necessary, EfiToolSys
> >>>>-
> >>>>-
> >>>>-class boottool(Grubby):
> >>>>- """
> >>>>- Client site side boottool wrapper.
> >>>>-
> >>>>- Inherits all functionality from boottool(.py) CLI app (lazily).
> >>>>- """
> >>>>- def __init__(self, path='/sbin/grubby'):
> >>>>- self.instantiated = False
> >>>>- self.path = path
> >>>>-
> >>>>-
> >>>>- def _init_on_demand(self):
> >>>>- if not self.instantiated:
> >>>>- try:
> >>>>- install_grubby_if_necessary()
> >>>>- Grubby.__init__(self, self.path)
> >>>>- self.instantiated = True
> >>>>- except Exception, e:
> >>>>- raise error.JobError("Unable to instantiate boottool:
> >>>>%s" % e)
> >>>>-
> >>>>-
> >>>>- def __getattr__(self, name):
> >>>>- self._init_on_demand()
> >>>>- return Grubby.__getattribute__(self, name)
> >>>>+from boottool_cli import Grubby as boottool,
> >>>>install_grubby_if_necessary, EfiToolSys
> >>>>
> >>>>This also seems to be working here.
> >>>Yes, exactly.
> >>>
> >>>But the lazy instantiation was done on purpose, and we may not be
> >>>able to get rid of it that easily. I'm doing some experimentation
> >>>here to think about the best approach.
> >>Lacking the historical context, what was the purpose originally?
> >Basically boiled down to avoiding triggering the grubby
> >build/installation code whenever
> >autotest.client.shared.boottool.boottool was instantiated. Code in
> >job.py instantiate boottool.boottool quite early.
>
> Now things look a lot clearer. If you look at:
>
> https://github.com/autotest/autotest/blob/master/client/job.py#L292
>
> you'll see boottool() is instantiated explicitly with the executable
> set to None.
>
> I'm posting a patches that, among other things, fix this.
>
> Please give it a try!
>
> Thanks!!!
Will revert the above diff and apply your 4 patches and let you know how
it goes.
-Nish
--
Nishanth Aravamudan <n...@us.ibm.com>
IBM Linux Technology Center
_______________________________________________
Autotest mailing list
Autotest@test.kernel.org
http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
