common/json.c: Check that JSMN result is well-formed.

[ZmnSCPxj]

xref: https://lists.ozlabs.org/pipermail/c-lightning/2020-June/000188.html

[ZmnSCPxj]

Probably deserves some kind of test though.

[ZmnSCPxj]

Modified the common/test/run-json.c, should be OK for merging now.

[cdecker]

Nice fix, just three tiny comments/clarifications needed.

Am I correct in assuming this is O(n) despite the recursive calls, since we advance the pointers by the parsed amounts in the recursive calls?

[ZmnSCPxj]

Am I correct in assuming this is O(n) despite the recursive calls, since we advance the pointers by the parsed amounts in the recursive calls?

Yes, it should be.

[ZmnSCPxj]

Am I correct in assuming this is O(n) despite the recursive calls, since we advance the pointers by the parsed amounts in the recursive calls?

Yes, it should be.

Ah, no, no. Deeply-nested objects/arrays would trigger the function being called over sub-ranges of an earlier call, so it does not achieve O(nd) where d is the nestedness of objects. Personally I think correctness is more important, we can always rewrite it faster later.

[ZmnSCPxj]

Sorry, completely rewrote the code to be O(n) on the number of tokens validated, and requiring O(d) stack space (d being nestedness of objects/arrays in the JSON input).

[ZmnSCPxj]

Renamed the validation function to validate_jsmn_parse_output.

[rustyrussell]

Ack f3438c2

[rustyrussell]

Nice work. Yes, I hit a similar issue previously, but that one was fixed upstream. Good to validate generally for this reason.