I have found XML repairing code in Python, and for some testing I did, it worked with a large output (20MB). But it will need to have better fault detection for some cases like XML element not complete.
About the development strategy, I think is better to create an independent tool and then incorporate with rebot. Would you like to share your Java code? We could translate to Python and/or expand to a better tool.
If @pekkaklarck approves this strategy, this is my proposal:

• The tool completes an XML output with missing closing elements.
• At the start (or near) of the XML fault, would introduce an output element with something like, "INFO: WARN: This report was recovered from a broken XML output!"
• Changes to rebot:
  1 - Capture first failure to process XML output and triggers repariring tool.
  2 - Process repaired XML output.
  3 - If still failed to process, stops with error (as is doing now)

I don't think Rebot should automatically try to fix outputs. We do, however, already have Fixml that does it: https://pypi.python.org/pypi/robotfixml.

Great!
So this feature request could be changed just to an improvement of the failure message at rebot.
It could give the error and recommend to install and use that tool.
Do you want me to propose a code change?

Mentioning Fixml in the error message when the output.xml is broken is a good idea.

I took a look at the tool and it uses BeautifulSoup 3 (which has been replaced by version 4). Tested with both versions (version 3 is slower, version 4 complaints about deprecated class). I have found a simpler tool, because robotfixml introduces one more dependency, is a Robot Framework output format specific, and has more complexity. A very first draft is at https://github.com/HelioGuilherme66/xml_repair_tool.

Yeah, Fixml is Robot specific on purpose. In addition to fixing xml it also tries to make it Rebot compatible.

It was created when there only was BeatifulSoup 3. Should update it to use v4 and also make it Python 3 compatible. I won't have time to look at it in the near future, but PRs are welcome.

I guess this issue can be closed. A separate issue and pr can be opened about mentioning Fixml in Rebot's invalid xml error message.

I don't think Rebot should automatically try to fix outputs. We do, however, already have Fixml that does it: https://pypi.python.org/pypi/robotfixml.

As far as this Rebot method should not autofix the output.xml file, fixing could easily be an optional argument with False as default.
I saw this robotfixml lib recently and not gonna lie... it made me laugh :P

@myjniak Nice for you to comment on a closed issue from 2015. Do you have any improvement to propose, in the form of code?