Concrete syntax for the LLVM language extension

[langston-barrett]

The only representation of Crucible programs using the LLVM language extension is as an in-memory data structure. Here are a few reasons we might want to have concrete, textual syntax for these programs:

• Tests of LLVM-related functionality are all currently integration tests involving llvm-pretty{,-bc-parser}. Not only is this not ideal from an isolation and modularity perspective, it is also slow.
• Furthermore, all of our tests have to decide between including binary blobs (.bc files) or depending on tools being present in the environment (clang or llvm-as), and executing those tools during the test suite (which is also unnecessarily slow).
• Textual test cases benefit from a sizable number of generic development tools (think fuzzers, delta debuggers, and tools like llvm-lit). With a some additional investment, very powerful tools like grammar-based fuzzers and test-case reducers become possible (e.g., llvm-reduce).
• Higher-level tools like SAW could cache Crucible-LLVM programs to disk for repeated/interactive use.

A reasonable outcome here (i.e., the ability to serialize reasonably-sized Crucible-LLVM programs that result from translation from LLVM modules) might depend on #74.

[bboston7]

What a coincidence! I'm currently working on adding support for arbitrary language extensions to Crucible Syntax. After that work is done you'll only need to write a small parser hook into Crucible Syntax for the LLVM extension.

[bboston7]

What a coincidence! I'm currently working on adding support for arbitrary language extensions to Crucible Syntax. After that work is done you'll only need to write a small parser hook into Crucible Syntax for the LLVM extension.

I just opened pull request #923 adding support for arbitrary language extensions.

[langston-barrett]

A bare-bones version of this is done, additional work can be tracked in other tickets (e.g., #1117, #1118).