Inconsistencies in data structures produced by Dyninst

[batuzovk]

As the result of parsing a binary file Dyninst produces several entities and some relations between them. Among entities there are functions, blocks, and jump tables. Both blocks and jump tables can belong to a function (or to several functions). These relations are described by multiple data structures that should be kept in consistent state, namely:

• For each function F from Block::getFuncs(), F should be in CodeObject::funcs() (i.e. it should 'exist' as a function).
• For each function F and block B, F should be in B.getFuncs() iff B is in F.blocks().
• For each jump table JT and function F, JT being in F.getJumpTables() should imply that JT.block is in F.blocks().

To Reproduce

1. Get attached program verify.cc.gz that tests the conditions above. Build it with g++ -std=c++0x -g verify.cc -I/path/to/dyninst/include -L/path/to/dyninst/lib -lparseAPI -linstructionAPI -lsymtabAPI -ldynDwarf -ldynElf -lcommon -lelf -ldwarf -o verify
2. Run on some sufficiently large guest binary file. Most common libraries produce at least some errors. libLLVM-17.so.1 for example, produces a lot.

Expected behavior
No inconsistencies found by the test program.

System (please complete the following information):

• Linux, x86_64 for both host and guest.
• Current master (commit cc5192e).

Additional notes
The most likely culprit is finalization. It changes function boundaries a lot, and I don't see a lot of code that fixes other data structures.

Also, block may not be part of the function that created it, but I'm not sure if this is intended or not.

[hainest]

This was fixed by #1668, no?

[batuzovk]

No? Why should it? It is a completely different issue. Reordering of functions doesn't affect those data structures.