init: fix unescape_string not advancing past escape characters

[nikromen]

After processing a \X escape sequence in the switch block, the read
pointer (val) was not advanced past the matched character. The next
loop iteration re-processed it as a literal, producing an
extra byte for every escape sequence.

For example, JSON "\n" decoded to {0x0A, 'n'} instead of {0x0A},
and JSON "\\" decoded to {'', }
instead of {''}.

Also fix the \u case: hexToDigit was called with a pointer to 'u'
instead of the first hex digit (val++ vs ++val), and the write index
was not advanced after memcpy, so decoded Unicode characters were
silently dropped. This caused e.g. \u0026 ('&') to be lost, breaking
any container command containing '&&'.

Reproducer (on fedora 43 x86):

[jkyjovsk@fedora libkrun] (fix-init-unescape-string) ↴
$ podman run --rm --runtime krun fedora:latest /bin/bash -c "printf 'hello\nworld\n'"
hello\
nworld\
n⏎

Expected output is:

hello
world

and second & bug:

[jkyjovsk@fedora libkrun] (fix-init-unescape-string) ↴
$ podman run --rm --runtime krun fedora:latest /bin/bash -c 'echo a && echo b'
a echo b

expected outoput:

a
b

[nikromen]

hmmm, both this and #668 are still not working as I thought :/ will look at it later 😴

[jakecorrenti]

Hi @nikromen, thanks for the contribution!

I just want to let you know this is also fixed in my latest draft PR porting the init code to Rust: #670