Ruby: stop considering post-update nodes to be local source nodes

[nickrolfe]

No description provided.

[nickrolfe]

Hmm, I may have got this wrong. On the one hand, dataflow queries on rapid7/metasploit-framework now complete in a reasonable time, but on the other hand, this doesn't look great:

Tuple counts for DataFlowPrivate::Cached::isLocalSourceNode#462ff392#f#antijoin_rhs@46f66bui:
        1463003991      ~0%    {2} r1 = JOIN boundedFastTC(DataFlowPrivate::Cached::localFlowStepTypeTracker#462ff392#ff_10#higher_order_body,DataFlowPrivate::Cached::TNode#462ff392#f) WITH DataFlowPrivate::Cached::TNode#462ff392#f ON FIRST 1 OUTPUT Lhs.1, Rhs.0
                           
            369909      ~3%    {1} r2 = JOIN r1 WITH DataFlowPrivate::Cached::entrySsaDefinition#462ff392#f ON FIRST 1 OUTPUT Lhs.1
                           
         731612184  ~88927%    {1} r3 = JOIN r1 WITH DataFlowPrivate::Cached::TExprNode#462ff392#ff_1#join_rhs ON FIRST 1 OUTPUT Lhs.1
                           
         731982093  ~81105%    {1} r4 = r2 UNION r3
                               return r4

[nickrolfe]

I still see the massive tuple count in the antijoin I posted above.

[hvitved]

I still see the massive tuple count in the antijoin I posted above.

I have pushed a fix for the antijoin.

[hmac]

@nickrolfe it sounds like Tom was happy with this change. I don't fully understand it but the test changes look plausible. Do you think after a rebase and DCA run we could merge it?

[nickrolfe]

@hmac the test changes I committed aren't the only ones – the n instanceof ExprNode change suggested by Tom introduced more test failures, and I haven't managed to look into them yet, but my gut instinct was that something got broken.

[hmac]

Ah OK - I was just checking in case this was something that was waiting on a final 👍 from someone.