Perl's syntax is undecidable, because the difference between treating some characters as a comment or as a regex can depend on the type of a variable that is only determined e.g. based on whether a search for a Collatz counterexample terminates, or just, you know, user input.
https://perlmonks.org/?node_id=663393
C++ templates have a similar issue, I think.
There’s a function called intuit_more which works out if $var[stuff] inside a regex is a variable interpolation followed by a character class, or an array element interpolation. Its result can depend on whether something in the stuff has been declared as a variable or not.
But even if you ignore the undecidability, the rest is still ridiculously complicated.
TIL! I went and dug up a citation: https://blog.reverberate.org/2013/08/parsing-c-is-literally-...
Parsing Bash is Undecidable - https://www.oilshell.org/blog/2016/10/20.html
I remember a talk from Larry Wall on Perl 6 (now Raku), where he says this type of thing is a mistake. Raku can be statically parsed, as far as I know.
Morbig: A static parser for POSIX shell - https://scholar.google.com/scholar?cluster=15754961728999604...
(at the time I wrote the post about bash, I hadn't implemented aliases yet)
But it's a little different since it is an intentional feature, not an acccident. It's designed to literally reinvoke the parser at runtime. I think it's not that good/useful a feature, and I tend to avoid it, but many people use it.