AdithyaSK's picture
AdithyaSK HF Staff
Add files using upload-large-folder tool
252bb4a verified
|
Raw History Blame Contribute Delete
2.15 kB

Fix the following issue:

Sync IDA Pro type definitions across collaborating analysts

We're using Polichombr to collaborate on reversing a sample with two other analysts. The current server-side IDA actions cover most of what we need — global names, comments, and structs all sync between our IDA instances through the API — but there's one IDA action that doesn't get shared: type definitions (the ones you set by hitting Y on a variable, function, or address in IDA Pro).

What we'd like to do

In IDA, I'll often refine a function prototype on a sub_xxxxxx, e.g. turn the default int __cdecl sub_401000(int, int) into something meaningful like BOOL __stdcall ParseConfigBlob(char *blob, size_t len). Then I want my teammates to pick that up automatically the next time their plugin pulls from the server, the same way they pick up the names and comments I made.

Right now, there's no way to do this through Polichombr. Names get pushed and pulled fine, structs get pushed and pulled fine, but the type info I set with Y just stays local in my .idb and the rest of the team has to re-discover it independently. For functions with non-trivial prototypes (lots of pointers, callbacks, custom typedefs) this ends up being a real source of duplicated work.

What we'd expect

Ideally the server should treat applied type definitions as just another kind of IDA action, alongside names / comments / structs:

  • the plugin should be able to push a type definition for a given sample at a given address,
  • and a teammate's plugin should be able to query, for a given sample, the type definitions that have been recorded — with the same kind of filtering you already support for names and comments (by address, or only the ones added after a given timestamp, so we don't keep re-pulling the whole history on every sync).

Would it be possible to add this? Happy to test against any branch.

(For API shape, mirroring the existing /samples/<sid>/names/ style would make sense — e.g. a types-flavored endpoint that accepts an address + typedef payload on push and returns the recorded typedefs on pull.)