Split binary format parsing into a shared crate

[Shnatsel]

I'm currently working on a project that requires very basic parsing of ELF, Mach-O and PE formats - essentially enumerating sections and extracting the data from a section with a specific name. libgoblin seems to be an overkill for this task, and I'd like to avoid unsafe code, so the parsers in cargo-bloat seem to fit the bill nicely.

Would you be interested in splitting the parsers into a shared crate? I'll harden the crate against untrusted input (panics, OOMs, etc), but that is probably not a concern for cargo-bloat. Perhaps wider testing of the shared code would be beneficial.

[RazrFalcon]

I'm not sure I have time to do this. Especially to support it later. But if you like, you can simply copy-paste the current code to your library and I will happily switch to it, as long is it doesn't have many dependencies (proc-macros in particular).

[Shnatsel]

Great! I will have to audit and support some code anyway, so I don't mind doing the work. I'll let you know once I have something usable.

[Shnatsel]

I've split the parsers into a shared library https://github.com/Shnatsel/binfarce, made the ELF parsers more general and hardened them against untrusted input so that they don't over-allocate or panic.

The changes this entails for cargo-bloat can be seen here: https://github.com/RazrFalcon/cargo-bloat/compare/master...Shnatsel:shared-library?expand=1 (although it still has a bit of file format detection code, I'm going to drop that too)

I'm open to feedback. If you don't have any objections, I'll give the Mach-o and PE parsers a similar treatment and open a pull request against cargo-bloat. I'm happy to add you as an owner on the git repo and on crates.io.

[RazrFalcon]

Yes, looking good.

[Shnatsel]

Mach-O is now converted too. I know section extraction works, but I couldn't test the full cargo-bloat workflow because I don't have a mac.

[Shnatsel]

The code looks pretty much complete to me now. I'll publish on crates.io and send a PR to cargo-bloat soon-ish.

I've refactored all 4 parsers, switched section extraction from allocations to iterators and callbacks, and fuzzed all 4 parsers. Removing allocations didn't affect the APIs used by cargo-bloat at all (symbol extraction still allocates).