[developers] Trouble Seeing Test Suite Instances

Joshua Crowgey jcrowgey at u.washington.edu
Thu Sep 11 23:22:34 CEST 2014 

Alec,

It is very mysterious about how the problem went away temporarily this
afternoon.  On my way home I realized that one thing we did differently
the time that things worked fine was that we didn't load the grammar
before trying to create tsdb instances.  I don't know why this would be
important or affect anything, but it is (as far as I can recall) the one
thing we did differently the time that we couldn't replicate the issue.

--Joshua

On 09/11/2014 01:59 PM, Alexander Sugar wrote:
> Joshua resolved the problem! Emily was right that it was a matter of
> where the database root was pointed, but not in the way I thought.
> Moving the /home or /tsdb/home folder(s) up or down before opening
> [incr tsdb()] and pointing the database root to that new location
> didn't have an effect. When we set the database root to the default
> /opt/delphin/lkb/src/tsdb/home, the instance showed up in the podium.
> After that, when we created in a new instance in the
> ~/Uyghur/tsdb/home location I had originally wanted, the instance
> appeared in the podium there as well.
> 
> -- Update, after getting home and restarting my computer, I am once
> again able to create a visible instance in the /opt/delphin/... path,
> but unlike earlier I was unsuccessful at doing so in my own folder.
> The source of this problem seems quite mysterious and I'd love to know
> what it is, but for now I'm happy to be able to create browsable
> instances somewhere and check coverage of my grammar.
> 
> Thank you,
> Alec
> 
> On 9/10/14, Alexander Sugar <sugara at uw.edu> wrote:
>> I'm seeing Joshua tomorrow am, so maybe he'll find the error somewhere in
>> my folders/files.
>>
>> Thanks again for looking at it, David.
>>
>> Emily M. Bender <ebender at uw.edu>于2014年9月10日星期三写道：
>>
>>> Thanks, David.  I think there is something in the interaction between
>>> [incr tsdb()] and UbuntuLkb---we saw something similar in 567 last year,
>>> but then it was possible to work around by just changing where in the
>>> directory tree the database root was.  Alec says that didn't help in his
>>> case, though, so I'm not sure this is the same thing.
>>>
>>> Emiliy
>>>
>>>
>>> On Wed, Sep 10, 2014 at 10:16 AM, CLMS help (David Brodbeck) <
>>> linghelp at uw.edu <javascript:_e(%7B%7D,'cvml','linghelp at uw.edu');>> wrote:
>>>
>>>> I tried booting Alec's VM with 'apparmor=0' set, and it made no
>>>> difference.  I'm also not seeing any "permission denied" type errors,
>>>> which
>>>> would be typical if AppArmor were the culprit.  I'm starting to suspect
>>>> an
>>>> issue with tsdb, since there are no error messages at all, but I don't
>>>> know
>>>> enough about its internals to diagnose it.
>>>>
>>>> On Mon, Sep 8, 2014 at 4:39 PM, CLMS help (David Brodbeck) <
>>>> linghelp at uw.edu <javascript:_e(%7B%7D,'cvml','linghelp at uw.edu');>>
>>>> wrote:
>>>>
>>>>> It's probably an AppArmor problem.  I thought I'd disabled it in recent
>>>>> builds but it may not have worked.  If you're comfortable changing
>>>>> kernel
>>>>> boot parameters, try adding 'apparmor=0' and see if that helps.
>>>>>
>>>>> On Mon, Sep 8, 2014 at 3:02 PM, Emily M. Bender <ebender at uw.edu
>>>>> <javascript:_e(%7B%7D,'cvml','ebender at uw.edu');>> wrote:
>>>>>
>>>>>> [Again, adding developers back to the cc.]
>>>>>>
>>>>>> Joshua, I think you just established that there's nothing wrong with
>>>>>> the skeleton, but that's not too surprising.  The problem seems to be
>>>>>> with
>>>>>> [incr tsdb()]'s ability to "see" certain directories on the UbuntuLKB
>>>>>> distro.  I'm cc-ing David B. in the hopes that he might have some time
>>>>>> to
>>>>>> look at this with Alec and/or some ideas.
>>>>>>
>>>>>> Emily
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Sun, Sep 7, 2014 at 9:33 AM, Alexander Sugar <sugara at uw.edu
>>>>>> <javascript:_e(%7B%7D,'cvml','sugara at uw.edu');>> wrote:
>>>>>>
>>>>>>> Joshua,
>>>>>>>
>>>>>>>      Somehow in updating my VirtualBox, I lost previous test suite
>>>>>>> instances. If I point my database root at the opt/delphin/... path
>>>>>>> that's set as a default, I am able to see those profiles in the
>>>>>>> podium
>>>>>>> after I create them (and I get the same "unable to write history file
>>>>>>> message", so that's definitely not the issue).
>>>>>>>
>>>>>>>     I'm using version 2.0 (16-jun-13; beta) of [incr tsdb()] on
>>>>>>> UbuntuLKB 17.
>>>>>>>
>>>>>>> - Alec
>>>>>>>
>>>>>>> On 9/6/14, Joshua Crowgey <jcrowgey at u.washington.edu
>>>>>>> <javascript:_e(%7B%7D,'cvml','jcrowgey at u.washington.edu');>> wrote:
>>>>>>>> Hi Alec,
>>>>>>>>
>>>>>>>> I doubt that .tsdb_history is the issue.  In Unix-like systems,
>>>>>>>> many
>>>>>>>> programs will store config info or other user-settings in files
>>>>>>>> that
>>>>>>>> start with "." in the user's home directory.  If you say (in a
>>>>>>> shell):
>>>>>>>>
>>>>>>>> $ ls -la ~/
>>>>>>>>
>>>>>>>> You'll see several such files, probably at least .emacs, .bashrc.
>>>>>>> In my
>>>>>>>> own system, I suppose that .tsdb_history stores your tsdb commands
>>>>>>>> so
>>>>>>>> that (maybe?) you can press "up" and see the previous command, or
>>>>>>>> something like that.  I found a .tsdb_history with these lines:
>>>>>>>>
>>>>>>>> jcrowgey at citrus:~$ cat .tsdb_history
>>>>>>>> do "/tmp/.tsdb.io.jcrowgey.11333.118378".
>>>>>>>> do "/tmp/.tsdb.io.jcrowgey.11333.118384".
>>>>>>>>
>>>>>>>> Because I've restarted my system since I last used tsdb, the /tmp
>>>>>>>> directory has been cleared and I'm not sure what was in those
>>>>>>>> /tmp/.tsdb.io.jcrowgey.* files.   My guess is those were a record
>>>>>>>> of
>>>>>>> my
>>>>>>>> sessions this morning when I was playing around with your item and
>>>>>>>> relations files.
>>>>>>>>
>>>>>>>> Anyway, the fact that the files were created is interesting, I
>>>>>>>> think.
>>>>>>>> It suggests that tsdb can create the new files, but we still don't
>>>>>>> know
>>>>>>>> why it can't read those files and display them in the podium.
>>>>>>>>
>>>>>>>> This makes me thing of another troubleshooting/debugging test.  Can
>>>>>>> you
>>>>>>>> see other existing testsuite instances?  That is, if you already
>>>>>>>> had
>>>>>>> a
>>>>>>>> profile with results and such, and you point your database root at
>>>>>>> that
>>>>>>>> dir, do those profiles show up in the podium?
>>>>>>>>
>>>>>>>> Sorry I don't really have good answers for you.  I just thought I'd
>>>>>>> jump
>>>>>>>> in and give you some stuff to try out.  By the way, are you using
>>>>>>>> the
>>>>>>>> knoppix+lkb distro?  What are the details of your platform?
>>>>>>>>
>>>>>>>> --Joshua
>>>>>>>>
>>>>>>>> On 09/06/2014 05:53 PM, Alexander Sugar wrote:
>>>>>>>>> Hi Joshua,
>>>>>>>>>
>>>>>>>>>       I think the reason I didn't see anything printed out in
>>>>>>>>> emacs
>>>>>>>>> was because I had previously 'created' an instance. When I cleared
>>>>>>> out
>>>>>>>>> the files inside of the home folder, I saw the same lines printed
>>>>>>>>> as
>>>>>>>>> you, except in the end:
>>>>>>>>>
>>>>>>>>> initialize(): creating empty data file for `score'.
>>>>>>>>> main(): no history file `.tsdb_history'.
>>>>>>>>>
>>>>>>>>> And of course no instance showed up in the podium. I'm guessing
>>>>>>>>> the
>>>>>>>>> '.tsdb_history' line is the source of the problem. Is there a
>>>>>>>>> quick
>>>>>>>>> fix for it? If so, that's great but I'm sorry to have put you
>>>>>>> through
>>>>>>>>> so much trouble just now...
>>>>>>>>>
>>>>>>>>> - Alec
>>>>>>>>>
>>>>>>>>> On 9/6/14, Joshua Crowgey <jcrowgey at u.washington.edu
>>>>>>> <javascript:_e(%7B%7D,'cvml','jcrowgey at u.washington.edu');>> wrote:
>>>>>>>>>> Hi Alec,
>>>>>>>>>>
>>>>>>>>>> I put your item and relations files into a dir and pointed my
>>>>>>> itsdb at
>>>>>>>>>> them and as far as I can tell, they loaded without a
>>>>>>>>>> problem---ie,
>>>>>>> tsdb
>>>>>>>>>> was able to create a new instance and I see it appear in the
>>>>>>>>>> podium
>>>>>>>>>> window.
>>>>>>>>>>
>>>>>>>>>> As to stderr, I think if you start tsdb in emacs then all of the
>>>>>>>>>> messages get printed there.  When I click to create a new
>>>>>>>>>> instance
>>>>>>> of
>>>>>>>>>> your item and relations files, I see:
>>>>>>>>>>
>>>>>>>>>> LKB(2): initialize(): creating empty data file for `analysis'.
>>>>>>>>>> initialize(): creating empty data file for `phenomenon'.
>>>>>>>>>> initialize(): creating empty data file for `parameter'.
>>>>>>>>>> initialize(): creating empty data file for `set'.
>>>>>>>>>> initialize(): creating empty data file for `item-phenomenon'.
>>>>>>>>>> initialize(): creating empty data file for `item-set'.
>>>>>>>>>> initialize(): creating empty data file for `run'.
>>>>>>>>>> initialize(): creating empty data file for `parse'.
>>>>>>>>>> initialize(): creating empty data file for `result'.
>>>>>>>>>> initialize(): creating empty data file for `rule'.
>>>>>>>>>> initialize(): creating empty data file for `output'.
>>>>>>>>>> initialize(): creating empty data file for `edge'.
>>>>>>>>>> initialize(): creating empty data file for `tree'.
>>>>>>>>>> initialize(): creating empty data file for `decision'.
>>>>>>>>>> initialize(): creating empty data file for `preference'.
>>>>>>>>>> initialize(): creating empty data file for `update'.
>>>>>>>>>> initialize(): creating empty data file for `fold'.
>>>>>>>>>> initialize(): creating empty data file for `score'.
>>>>>>>>>> close-connection():
>>>>>>>>>>
>>>>>>> `/home/jcrowgey/Desktop/alexsugar/tsdb/anonymous/alex/14-09-06/lkb1/'
>>>>>>>>>> expiry
>>>>>>>>>>
>>>>>>>>>> You don't see anything happening there, eh?  I'm not sure it
>>>>>>>>>> would
>>>>>>> help,
>>>>>>>>>> but it might be interesting to try this:
>>>>>>>>>>
>>>>>>>>>> Since all the tsdb instance is (before you use it) is just a set
>>>>>>>>>> of
>>>>>>>>>> empty files + copies of the item and relations files, you could
>>>>>>> simply
>>>>>>>>>> make the instance yourself on the command line and then see if
>>>>>>> tsdb can
>>>>>>>>>> recognize and already created one.  Perhaps this would give a
>>>>>>>>>> clue
>>>>>>> as to
>>>>>>>>>> whether the issue is in creating files or if reading files is
>>>>>>> affected
>>>>>>>>>> too.
>>>>>>>>>>
>>>>>>>>>> I think the usual directory structure is:
>>>>>>>>>>
>>>>>>>>>> $DATABASEROOT/NAME_OF_GRAMMAR/NAME_OF_TS/DATE/lkb/
>>>>>>>>>>
>>>>>>>>>> Then, inside that final dir, you just put a copy of item and
>>>>>>> relations
>>>>>>>>>> plus empty files with the following names:
>>>>>>>>>>
>>>>>>>>>> jcrowgey at citrus:~/Desktop/alexsugar/tsdb$ ls -la
>>>>>>>>>> anonymous/alex/14-09-06/lkb
>>>>>>>>>> total 56
>>>>>>>>>> drwxr-xr-x 2 jcrowgey jcrowgey  4096 Sep  6 13:14 .
>>>>>>>>>> drwxr-xr-x 4 jcrowgey jcrowgey  4096 Sep  6 13:15 ..
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 analysis
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 decision
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 edge
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 fold
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey 36234 Sep  6 13:14 item
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 item-phenomenon
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 item-set
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 output
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 parameter
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 parse
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 phenomenon
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 preference
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey  9746 Sep  6 13:14 relations
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 result
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 rule
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 run
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 score
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 set
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 tree
>>>>>>>>>> -rw-r--r-- 1 jcrowgey jcrowgey     0 Sep  6 13:14 update
>>>>>>>>>>
>>>>>>>>>> I don't know if it will help, but if you can initialize your own
>>>>>>>>>> instance and tsdb is able to read it, that might help you go
>>>>>>>>>> ahead
>>>>>>> with
>>>>>>>>>> your grammar testing and it might help "developers" get an idea
>>>>>>>>>> of
>>>>>>> what
>>>>>>>>>> part of the program is going wrong.
>>>>>>>>>>
>>>>>>>>>> Sorry I don't have anything better for you,
>>>>>>>>>>
>>>>>>>>>> --Joshua
>>>>>>>>>>
>>>>>>>>>> On 09/05/2014 05:01 PM, Alexander Sugar wrote:
>>>>>>>>>>> I don't see any error messages appearing either in STDERR (which
>>>>>>>>>>> I
>>>>>>>>>>> assume is the bar at the bottom-right of the [tsdb()] window) or
>>>>>>> emacs
>>>>>>>>>>> (which I used to start tsdb).
>>>>>>>>>>>
>>>>>>>>>>> I'm attaching my 'item' and 'relations' files. Thanks a lot for
>>>>>>> looking
>>>>>>>>>>> at
>>>>>>>>>>> them.
>>>>>>>
>>>>>>>>>>> - Alec
>>>>>>>>>>>
>>>>>>>>>>> On 9/5/14, Joshua Crowgey <jcrowgey at u.washington.edu
>>>>>>> <javascript:_e(%7B%7D,'cvml','jcrowgey at u.washington.edu');>> wrote:
>>>>>>>>>>>> I see.  In fact, now that I think about it, the fact that your
>>>>>>>>>>>> testsuite
>>>>>>>>>>>> appears under File > Create seems to indicate that skeletons
>>>>>>>>>>>> and
>>>>>>>>>>>> Index.lisp are correct.
>>>>>>>>>>>>
>>>>>>>>>>>> When you click on your testsuite under File > Create and
>>>>>>>>>>>> nothing
>>>>>>>>>>>> appears
>>>>>>>>>>>> in the window, is there any text or error messages that appear
>>>>>>>>>>>> in
>>>>>>>>>>>> STDERR
>>>>>>>>>>>> or in emacs (if you started tsdb using the emacs interface)?
>>>>>>>>>>>>
>>>>>>>>>>>> Also, if you want to send me your 'item' and 'relations' files
>>>>>>>>>>>> I
>>>>>>> can
>>>>>>>>>>>> try
>>>>>>>>>>>> to replicate your error on my system.
>>>>>>>>>>>>
>>>>>>>>>>>> --Joshua
>>>>>>>>>>>>
>>>>>>>>>>>> On 09/05/2014 03:37 PM, Alexander Sugar wrote:
>>>>>>>>>>>>> Hi Joshua,
>>>>>>>>>>>>>
>>>>>>>>>>>>>      Thanks for offering your help. I just double checked, and
>>>>>>> my
>>>>>>>>>>>>> skeletons directory indeed has Relations (with a capital 'R')
>>>>>>> and
>>>>>>>>>>>>> Index.lisp. The latter points to the folders 'testsuite' and
>>>>>>>>>>>>> 'testcorpus', which each contain an 'item' and 'relations'
>>>>>>>>>>>>> file.
>>>>>>>>>>>>>
>>>>>>>>>>>>> - Alec
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 9/5/14, Joshua Crowgey <jcrowgey at u.washington.edu
>>>>>>> <javascript:_e(%7B%7D,'cvml','jcrowgey at u.washington.edu');>> wrote:
>>>>>>>>>>>>>> Hi Alex,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I'm not super-fluent with tsdb, but just to make sure you
>>>>>>> tried it:
>>>>>>>>>>>>>> have
>>>>>>>>>>>>>> you double-checked that the skeletons root is set correctly?
>>>>>>> It
>>>>>>>>>>>>>> should
>>>>>>>>>>>>>> be pointing to a directory that contains Relations and
>>>>>>> Index.lisp
>>>>>>>>>>>>>> where
>>>>>>>>>>>>>> Relations gives the schema for the fields in the tsdb
>>>>>>>>>>>>>> profiles
>>>>>>> and
>>>>>>>>>>>>>> Index.lisp gives paths to testsuites.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> --Joshua
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> On 09/05/2014 12:01 PM, Alexander Sugar wrote:
>>>>>>>>>>>>>>> To whom it may concern,
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>      I am a graduate student at the University of Washington
>>>>>>>>>>>>>>> learning
>>>>>>>>>>>>>>> how to use Delph-in tools like [incr tsdb()]. Recently, when
>>>>>>> I try
>>>>>>>>>>>>>>> to
>>>>>>>>>>>>>>> process my small uig grammar in [incr tsdb()], I see my test
>>>>>>> suite
>>>>>>>>>>>>>>> and
>>>>>>>>>>>>>>> one test corpus under File > Create. After clicking
>>>>>>>>>>>>>>> 'create',
>>>>>>>>>>>>>>> however,
>>>>>>>>>>>>>>> no test suite instances appear in my window. Could you
>>>>>>>>>>>>>>> please
>>>>>>> help
>>>>>>>>>>>>>>> me
>>>>>>>>>>>>>>> fix this problem?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Thank you,
>>>>>>>>>>>>>>> Alexander Sugar
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Emily M. Bender
>>>>>> Associate Professor
>>>>>> Department of Linguistics
>>>>>> Check out CLMS on facebook! http://www.facebook.com/uwclma
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> David Brodbeck
>>>>> System Administrator, Linguistics
>>>>> University of Washington
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> David Brodbeck
>>>> System Administrator, Linguistics
>>>> University of Washington
>>>>
>>>
>>>
>>>
>>> --
>>> Emily M. Bender
>>> Associate Professor
>>> Department of Linguistics
>>> Check out CLMS on facebook! http://www.facebook.com/uwclma
>>>
>>
>

More information about the developers mailing list