In addition to treating comparisons with literals as sanitisers, we now
also treat comparisons with variables that have a single assignment as
sanitisers.
Proving that such a variable is actually a constant is not easy, but for
this use case a simple approximation works fine.
`SEMMLE_TYPESCRIPT_NODE_RUNTIME` can be used to provide the path to the Node.js runtime executable.
If this is omitted, the extractor defaults to the current behaviour of looking for `node` on the PATH.
`SEMMLE_TYPESCRIPT_NODE_RUNTIME_EXTRA_ARGS` can be used to provide additional arguments to the
Node.js runtime. These are passed first, before the arguments supplied by the extractor.
These changes are designed to allow TypeScript extraction in controlled customer environments where
we cannot control the PATH, or must use custom Node.js executables with certain arguments set.
This test case exposes two bugs in our data flow library (fixed by the
two previous commits):
- the charpreds of `SourcePathNode` and `SinkPathNode` only ensured
that they were on a path from a source to a sink, not that they
actually were the source/sink themselves;
- function summarization would allow for non-level paths; in the
test case, this meant that one of the summaries for `source`
represented the path returning from `source` on line 13 and then
flowing back into the call on line 15, in the process transforming
the parity of the flow label and hence causing a spurious flow.
Their charpreds previously only ensured that they were on a path from a
source to a sink, not that they actually were the source and sink,
respectively. See two commits further for a test case.
It now also has `step` and `smallstep` predicates. In the usual case,
however, I think I prefer the `SourceNode::track` API, so I left the
recommended style in the qldoc alone (and adjusted the one for
`TypeBackTracker` to match).
