jsonnetfmt panic when encountering unescaped regex in string

[mihaitodor]

This issue was reported originally on the Tanka repo here, but it can be easily reproduced with jsonnetfmt.

Given this invalid jsonnet:

{
  x: {
    data: '([^:]+)(?::\d+)?',
  },
}

running jsonnetfmt test.jsonnet (latest main - 51daeb3) panics with the following stack:

panic: Badly formatted string, should have been caught in lexer.

goroutine 1 [running]:
github.com/google/go-jsonnet/internal/formatter.(*EnforceStringStyle).LiteralString(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000cc210, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/formatter/enforce_string_style.go:45 +0x1e5
github.com/google/go-jsonnet/internal/pass.(*Base).Visit(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000f8170, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:431 +0x297
github.com/google/go-jsonnet/internal/pass.(*Base).ObjectField(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000f80f0, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:160 +0x193
github.com/google/go-jsonnet/internal/pass.(*Base).ObjectFields(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000cc330, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:191 +0x76
github.com/google/go-jsonnet/internal/pass.(*Base).Object(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000cc2c0, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:331 +0x66
github.com/google/go-jsonnet/internal/pass.(*Base).Visit(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000f8350, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:435 +0x8a3
github.com/google/go-jsonnet/internal/pass.(*Base).ObjectField(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000f82d0, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:160 +0x193
github.com/google/go-jsonnet/internal/pass.(*Base).ObjectFields(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000cc3e0, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:191 +0x76
github.com/google/go-jsonnet/internal/pass.(*Base).Object(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc0000cc370, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:331 +0x66
github.com/google/go-jsonnet/internal/pass.(*Base).Visit(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc000096650, 0x0, 0x0)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:435 +0x8a3
github.com/google/go-jsonnet/internal/pass.(*Base).File(0xc0000d8060, 0x11ef0e0, 0xc0000d8060, 0xc000096650, 0xc0000a6090)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/pass/pass.go:461 +0x88
github.com/google/go-jsonnet/internal/formatter.visitFile(...)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/formatter/jsonnetfmt.go:133
github.com/google/go-jsonnet/internal/formatter.Format(0x7ffeefbffacb, 0xc, 0xc0000d8000, 0x2e, 0x2, 0x2, 0x1, 0x1, 0x1010001, 0x89fffff, ...)
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/internal/formatter/jsonnetfmt.go:168 +0x5f6
main.main()
	/Users/mihaitodor/Projects/go/pkg/mod/github.com/google/[email protected]/cmd/jsonnetfmt/cmd.go:283 +0x9c5

While jsonnetfmt is correct to reject this malformed code, panicking is not ideal.

[sbarzowski]

Thanks for reporting.

The panic happens, because the transformations conceptually cannot fail. They should never be given bad input in the first place.

The easiest fix would be to add a dry run of parser.StringUnescape in the lexer:

• https://github.com/google/go-jsonnet/blob/v0.17.0/internal/parser/lexer.go#L887
• https://github.com/google/go-jsonnet/blob/v0.17.0/internal/parser/lexer.go#L905

[mihaitodor]

Thank you for looking into this @sbarzowski! I haven't dug that deep into the code yet, but I did notice that the jsonnet command somehow handles this gracefully, so my initial guess was that jsonnetfmt is missing some preflight check.

jsonnet test.jsonnet
<stdin>:3:11-29 Unknown escape sequence in string literal: \d

    data: '([^:]+)(?::\d+)?',

I might try to take a stab at sending a PR, but no promises! :)

[netomi]

This also seems to be a duplicate of #606