Bug #21849: Ripper.lex emits two `tstring_content` for single string in edge-case

• Status: Open
• Assignee: -
• Target version: -
• ruby -v: ruby 4.0.0 (2025-12-25 revision 553f1675f3) +PRISM [x86_64-linux]
• Backport: 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN, 4.0: UNKNOWN

[Earlopain (Earlopain _)]

Ripper.lex('"x#$%"')
=> 
[[[1, 0], :on_tstring_beg, "\"", BEG],
 [[1, 1], :on_tstring_content, "x", BEG],
 [[1, 2], :on_tstring_content, "\#$%", BEG],
 [[1, 5], :on_tstring_end, "\"", END]]

It seems like ripper sees #$ and assumes that this is bare interpolation. However, $% is not a valid global name.
So I think it should instead be a single tstring_content:

[[[1, 0], :on_tstring_beg, "\"", BEG],
 [[1, 1], :on_tstring_content, "x\#@%", BEG],
 [[1, 5], :on_tstring_end, "\"", END]]

I stumbled across this in the prism translator for ripper. Prism only emits a single token for the string.