Unicode 'BACKSPACE' stop to read project

[thisandagain]

Expected Behavior

Should be open and run project in v3 as v2.

Actual Behavior

If a project contains Unicode character 'BACKSPACE' (\u0008), it will not run on Scratch 3.0 beta. The scripts are disappeared after the character and not editable in v3.0 any more.

Steps to Reproduce

Unicode Character 'BACKSPACE' (\u0008)

Unicode Character 'BACKSPACE' (\u0008) on Scratch 3.0 beta

Operating System and Browser

MacOS 10.13.6
Chrome 70.0.3538.110

---

Moved from scratchfoundation/scratch-gui#4190
/cc @yokobond

[thisandagain]

/cc @ktbee @kchadha

[kchadha]

This is directly related to #1077. The problem with both of these is that xml does not allow control characters.

One of the problems I was dealing with when I was trying to resolve this issue a while back was -- what is/are the place(s) where we should catch and fix this issue since these illegal characters can sneak into the block XML from a variety of different places (listing just a few below):

• imported project (e.g. from 2.0)
• variable name via the variable modal -- how likely is it that this character can be typed unintentionally...
• a different text field in the 3.0 editor whose contents can appear on a block (e.g. sprite name, costume/sound name, etc.)

[yokobond]

The cause of this problem is that the error 'PCData invalid Char value' was returned by DOMParser when making svg text from string in the project at SvgRenderer.loadString(svgString, fromVersion2).

Invalid char like that is not only \u0008 but also \u001c \u001d etc. (ref: https://www.w3.org/TR/xml/#charsets ). These invalid characters can be easily and unconsiously input by the default Japanese IME on MacOS, then so many project was destructed in Japanese and will be happen in the future continuously.