As discussed with @hvitved offline. This helps out to ensrue we don't
needlessly evaluate dataflow for configurations that are not needed
anyway. That is, if other library modeling also used the same dataflow
configuration, which ends up being used in query A, then dataflow for
all the `DataFlowImplForLibraries` configurations would be computeted at
once. When we get to evaluate the query `RequestWithoutValidation.ql`
these results mgith have been forgotten since the predicates are not
cached, and everything will have to be computeted again.
In principle we could be added a dataflow copy for each framework.
However, since we know that the `disablesCertificateValidation`
member-predicates for all the HTTP client libraries will all be used at
the same time, and only for the one query, we only add ONE additional
copy.
Note that the only use of `DataFlowImplForLibraries` before this PR is
using `tainttrackingforlibraries.TaintTrackingImpl` (based on
DataFlowImplForLibraries) for regex computation.
c904ba1d16/ruby/ql/lib/codeql/ruby/Regexp.qll (L153)
Since this is currently transitively imported from Frameworks.qll
(through Core.qll, and core/String.qll), the previous approach didn't
actually violate the assumption about all configurations always being in
scope, but it might have been more by accident, than by purpose.
Most of the classes and predicates in this query can be shared between
the two languages. There's just a few language-specific things that we
place in IncompleteMultiCharacterSanitizationSpecific.
With this change, users are now able to run View AST command in
vscode within vscode workspaces that do not include the core libraries.
The relevant core library only needs to be installed in the package
cache.
By explicitly marking children in the `schema.yml` file, an internal
`getAChild` predicate is implemented, that is in turn used in `AstNode`
to implement `getParent`.
This is yet to be used in the control flow library to replace the
hand-rolled implementation.
A further, more complex step is to use the same information to fully
generate the core implementation of `PrintAst` (including the
accessor string). This will be done later.
The `parent` tests use the same swift code as the extractor tests, and
this is currently enforced by `sync-files.py`. Notice that `qltest.sh`
had to be modified to deal with multiple files, which was not working
yet.