Mach-O: Add the ability to parse the dyld exports trie

[PaulDance]

Mach-O objects can contain a dyld info command that points to various data structures. Among those is the exports trie that stores the set of symbol names to be exported upon loading by dyld, if I understand things correctly. This particular structure can also be accessed to by looking for the LC_EXPORTS_TRIE load command.

Parsing such data is really not trivial, as it involves recursing through nodes and edges, so having it in object would be nice. Currently, the mach_object crate offers a correct implementation: I've been using it for a work project and it seems to behave well. However, this crate does not seem to be maintained anymore and the maintainer seems unreachable. We would therefore gladly work on porting the functionality to object instead and therefore upstream part of our work.

A few questions about the contribution then:

• Does this even have its place in object?
• What would go in macho vs. read::macho? There aren't much definitions similar to the rest of macho, it's mostly the parsing algorithm that needs implementation.
• Would it be desirable to add it to a higher-level API?
• If so, should it be added directly onto MachOFile or to the Object trait and its implementation of it? Having it outside of the trait makes more sense to me because it would not be behavior common to all formats, but I'm just checking anyways.

[philipc]

These three issues all seem reasonable, but I'm not at all familiar with these parts of Mach-O. A starting point would be to modify the readobj example to be able to display these data structures. This would be primarily using the low level API accessed via MachHeader, rather than MachOFile or Object. Adding a method to MachOFile too would be fine, but they probably don't belong in Object.

macho contains the raw file structures and constants. These are the things that you would find in a specification for the file format, and typically come from C header files. read::macho contains the code to do the parsing. This can include methods for the structures in macho.

[PaulDance]

Thanks for detailed answer! It is indeed consistent with what I had in mind. I should therefore be able to start working on this, some time soon hopefully.