Further fftb woes

Following up on my previous post, I’m trying fftb on the latest Ubuntu+LKB VM and running into various problems.

1: I have a profile that was produced by the aggregation evaluation set up, and with fftb pointed at the profile, I can get the browser window, but it shows “404 no stored forest found for this item” for every item. So, I’m guessing that there’s an incompatibility with the profile.

2: I used mkprof to create clean profile and then tried to use art+ace as documented here to process it:

… but I get 0 parses for every item.

3: I tried to then use the corresponding ace version over on my Mac OS to at least produce the profile, but I get errors like this:

erb@nindra:~/tmp/agg-fftb$ art-0.1.9/art -f -a '/Users/erb/tmp/agg-fftb/ace-0.9.31/ace --disable-generalization -g abz1/ace/abz1.dat -O' abz1fftb
reading results for                0	out of sync; arbiter sent 'dlopen: dlopen(/var/folders/99/ghgtffc97zxct161zd0hfscc0000gn/T//qcex.so-DX3shF, 2): no suitable image found.  Did find:' when expecting a SENT or SKIP header.
failed to read result for 0

I tried just parsing with ace (without art) and saw something similar:

erb@nindra:~/tmp/agg-fftb$ ace-0.9.31/ace -g abz1/ace/abz1.dat 
la teitu yo he-iya yo fu munuma
dlopen: dlopen(/var/folders/99/ghgtffc97zxct161zd0hfscc0000gn/T//qcex.so-tZxd6L, 2): no suitable image found.  Did find:
	/var/folders/99/ghgtffc97zxct161zd0hfscc0000gn/T//qcex.so-tZxd6L: unknown file type, first eight bytes: 0x7F 0x45 0x4C 0x46 0x02 0x01 0x01 0x00
	/private/var/folders/99/ghgtffc97zxct161zd0hfscc0000gn/T/qcex.so-tZxd6L: unknown file type, first eight bytes: 0x7F 0x45 0x4C 0x46 0x02 0x01 0x01 0x00

Parsing like that at the command line on linux does give me results, so that’s the not the source of the problem over there.

Any pointers greatly appreciated…

For completeness, here is how I’m invoking art+ace on linux and the first bit of results:

art-0.1.9/art -f -a '/home/ubuntu/agg-fftb/ace-0.9.31/ace --disable-generalization -g abz1/ace/abz1.dat -O' abz1fftb/
reading results for                0    0 results
reading results for               10    0 results
reading results for               20    0 results

But on interactive parsing, I can verify that item 0 actually does parse. I suspect that there is some error in now I’m calling ace which is getting suppressed by art…

I’m not certain, but just to eliminate possibilities…

The “Trouble Shooting” section of the FftbTop page suggests that, for this error message, there are either missing lexical entries or tokenization didn’t happen correctly. Maybe check if the REPP module is loaded correctly? Although this would be surprising if you’re able to parse directly with ACE using the same grammar image.

Are you thinking that it didn’t parse because it says “0 results” here, or because the edge files were empty? I think art reports “0 results” because the full-forest parsing doesn’t enumerate all analyses. For instance, I created a profile with 2 sentences, and here’s what I see:

$ art -f -a 'ace -g erg-2018-x86-64-0.9.34.dat -O' foo
reading results for                1	0 results
reading results for                2	0 results
$ wc -l foo/edge
25 foo/edge
$ head -1 foo/edge
845@1@root_strict@4@1@0@3@@(843)@@

To confirm: 0 results reported by art is not a problem in this scenario. Did you try running FFTB with the profile you created under Linux? The art invocation looked plausible to me.

Regarding the MacOS error you saw, certainly the art command is failing because of an underlying problem ace is encountering. I can’t tell for sure what it is, but at a guess: Is this by any chance on Apple silicon? Also, is that .dat grammar image compiled on the same computer and using the same version of ace?

The 404 error you saw from FFTB most likely means the profile in question was parsed for ordinary trees instead of for full forest, if memory serves. That would seem a likely explanation if the profile you tried on came out of the aggregation internal evaluation workflow.

Thank you, @goodmami and @sweaglesw ! Indeed, the linux profile did work with fftb — I just was fooled by the string of ‘0 results’ messages.

For the problem on MacOS, I’m not quite sure where I was in versions, but this is not an M1 machine. OTOH, that was probably a false start and not worth pursuing further.

So, I’m set up on the Ubuntu+LKB VBA, with fftb working!

Still: are there any plans to produce compiled versions of fftb for MacOS? I ask because the Ubuntu GUI is very clunky in my set up…