Hi I'm getting an exception when trying to use the torecord() fun

Are you testing the function like this in isolation? Or is it being used as part of a larger XML document? If you have XML headers, try removing the ones such as:
<?xml version="1.0" encoding="UTF-8"?>
<Instruction xmlns="http://tempuri.org/">

I am testing the function with the xml fragment as shown, so there are no headers.

I was able to reproduce the error in 7.4. However, when testing in 7.5 and 7.6, this function with the special character gives the result as ?g, rather than erroring out. This is a bit more graceful, so we know the behavior has been updated in later versions of Appian.

The root issue seems to be that the document isn't encoded in UTF-8, even though it is being read as if it were. There might be a limitation on this working by itself (as you are testing in the expressions editor), but we have seen cases where this works properly with special characters when populating a CDT in a full XML document, and stripping out the headers as I mentioned above.

Thanks Aleski.
The error can be reproduced outside of the expressions editor.
There does seem to be some mis-handling of the character set encoding within Appian. The xml fragment is encoded as UTF-8 originally, but this seems to be re-encoded as ISO-8859-1 and then treated as UTF-8 which causes the error.

When you reproduce it outside of the expressions editor, is it still just as a standalone function? What is the use case for using it like this?

The actual problem is that we ase passing UTF-8 encoded XML from the database and Appian is re-encoding it as ISO-8859-1. I have raised a support case.

I see that we have provided a workaround in that case. For reference, it is to use something like:
torecord("<?xml version='1.0' encoding='ISO-8859-1'?><tag>ég</tag>", "")

With Appian Support's help, we have configured JBoss JVM settings with UTF-8 as the default file encoding. This step is missing in the online documentation.