dtbook-to-odt hangs with math

[egli]

When translating a fairly complicated Math book the conversion hangs, see https://gist.github.com/egli/82d086ed9a0fefc92394541dd4ff1546. It hangs regardless of whether it is invoked via Daisyproducer or directly via the command line. If math is disabled it works like a charm.

[egli]

Bert suspects that this could be a problem with the asciimath handling. This is a piece of Javascript code and in Java8 the Javascript engine has changed:

about the asciimath thing see what the difference is between java 7 and java 8 in terms of javascript engine and, if it is like i suspect, namely that 7 uses rhino and 8 nashorn then maybe you have to explicitly use rhino? i found an article about it: https://wiki.openjdk.java.net/display/Nashorn/Using+Rhino+JSR-223+engine+with+JDK8

[egli]

When running the pipeline2 from the command line with

su pipeline2 -s /opt/daisy-pipeline2/bin/pipeline2

and then converting with

eglic@xmlp01:~$ /opt/daisy-pipeline2-cli/dp2 sbs:dtbook-to-odt --source /tmp/daisyproducer-3H6EW7/30309.xml --images DROP --asciimath BOTH --output test-results2

it produces an odt file.

Also when invoking the thing via the web interface (connecting to the pipeline2 that was started like above) an odt file is returned

[egli]

OK, the whole thing also works when invoking the standard way, i.e. via Daisyproducer Web Interface and the pipeline2 has been started via sv start daisy-pipeline.

I'm using the book in question (Mathematik 3. Arbeitsheft II), the same that was used in the gist.

It works for all three math options: Both, AsciiMath and MathML

[egli]

In other words: Looks like I can no longer reproduce the problem

[egli]

Haha, the asciimath was commented out in the dtbook-to-odt module. No wonder it worked

[egli]

I extracted the formulas in a gist

[bertfrees]

@egli I created a unit test in asciimath-utils, but I can not make it fail on my Debian container (Stretch + openjdk-8-jdk package). I also tried running the whole book.

[egli]

see https://gist.github.com/egli/54367184ebf4c80fbbdf26339cae2b9f for another extensive log

[egli]

The Unit Tests pass on the server machine, so maybe it isn't a problem with asciimath after all.

And consequently switching Rhino for Nashorn wouldn't even help

[egli]

The problem was actually a deadlock in PushNotifier.java

[egli]

Fixed in sbsdev/pipeline@86537c75a9a1fdf74cdb2bb178f09792336fd997

[egli]

@bertfrees would you mind merging this stuff into the sbs branch? Thanks

[bertfrees]

Done