[incr tsdb()] podium consistently dying on a test suite

What could be the reason for [incr tsdb()] podium to consistently “die” (display a cross instead of a normal mouse pointer, for an indefinite time) after what seems like an otherwise successful processing of a short test suite? It happens often but not every time on this particular test suite and very rarely on other test suites that I have, I think.

Update: I deleted the test suite, created it again, and I think this stopped happening for now.

Further update: Nope, still happening :(.

The fact that it only happens “sometimes” makes me wonder if it’s a memory issue. Is there a rule that could be spinning? If you try processing with pydelphin instead of [incr tsdb()] you might get an error message or warning that’s informative.

1 Like

This sounds like a question for Stephan, but I don’t think he checks this site (at least he is not a registered user that I can see). You might have better luck taking this to the developers list.

You could try parsing your test suite directly in the LKB, via the Parse -> Batch parse… menu command. This understands the format of [incr tsdb()] ‘item’ files.

1 Like

@johnca, are you suggesting that I would then see some error messages? I am not seeing any in the terminal, it seems like everything I expect to be parsed is successfully parsed, etc. But I might try and see if I spot something.

Should we submit the item file or the profile directory?

@olzama, yes, any error messages should appear in the terminal. It sounds as though parsing per se isn’t the problem, in which case the problem probably lies somewhere in [incr tsdb()] - either on the Lisp or the tcl/tk side. I saw the cross cursor a few times when getting [incr tsdb()] to work in lkb-fos, but didn’t investigate because I had been doing some pretty invasive things to the system.

@arademaker, yes, submit the item file. But I don’t recommend it as part of a normal [incr tsdb()] workflow since the output is just a vanilla text file, not a proper profile.

1 Like

Here’s the terminal run which resulted in the podium dying. Nothing untoward, except for the lexical analysis error at the end, but that shouldn’t lead to the podium dying, should it?

Screen Shot 2020-06-17 at 5.08.03 PM

The terminal output looks perfectly OK, so no clues there. The only potential cause I can think of is a Unicode encoding/decoding problem in communicating with swish++. Could you email me the grammar and test suite so I can try to reproduce this podium lock-up? Thanks.