Invalid command line xml output (COMPOUND)

[kijerk]

After upgrading from LT 4.8 to 5.0 there are (depending on the input) sometimes lines like COMPOUND: ... at the beginning of the xml output, which render the file invalid for further processing.

For example:

COMPOUND: Kühlschmier-mittel
COMPOUND: Wartungs-zugang
COMPOUND: Gliedmassen
COMPOUND: Gliedmassen
COMPOUND: Klettbänder
COMPOUND: Wechslerarms
COMPOUND: Destillatgleiches
COMPOUND: Herunterfahrens
COMPOUND: Wechslergabel
COMPOUND: Spindelkopf
COMPOUND: Spannzangenschlüssel
<?xml version="1.0" encoding="UTF-8"?>
...

Can you please remove the starting COMPOUND: ... lines from the output? I do not know if this was introduced with version 4.9 or 5.0. Switching to 4.8 does not produce these lines. Thank you.

Just another output from a different file. Hope it helps:

COMPOUND: Lagerfreundliches
COMPOUND: Gießharzsystem
COMPOUND: Umgebungs-temperatur
COMPOUND: Umgebungs-temperatur
COMPOUND: Strangverlauf
COMPOUND: Strangverlauf
10:33:34.375 [ForkJoinPool-1-worker-2] INFO org.languagetool.rules.de.GermanSpellerRule - UNKNOWN: hochbeständigen
COMPOUND: Hochwärmebeständige
COMPOUND: Strangführung
COMPOUND: Strangführung
COMPOUND: Strangführung
COMPOUND: Strangverlauf
...

[danielnaber]

We use these to find words not known to the dictionary. Quick workaround: filter out lines that match ^COMPOUND: before interpreting as XML.

[kijerk]

This is the kind of workaround I had in mind, too. But it is just a workaround. I don't know exactly what else can be in front of the real xml content. And as you can see, the line 10:33:34... doesn't start with COMPOUND:.

The lines are no xml and file is not valid. Why is this kind of output in the xml file? I would really like if you could try to fix the source of the problem.

[kijerk]

@danielnaber Is there a chance for a fix from your side? It would be really unfortunate, if the xml-output would not work by default...

[danielnaber]

The fix is easy, but so far there are no plans to have a 5.0.1 because of this bug, so the fix would be in the snapshots only.

[kijerk]

Great, good news and good to know!

[valentjn]

This breaks all tools which use the Java interface, but depend on a clean stdout (e.g., language servers where stdout is directly connected to a client via RPC). An ugly workaround is to silence LanguageTool with https://stackoverflow.com/q/4799006.