"Create Function Here" has weird ARM vs Thumb behavior

[percontation]

Setup:

1. I have a small raw binary, it's mostly thumb instructions
2. I mark some functions as thumb, so Binary Ninja now shows the default architecture as thumb.
3. There are additional chunks of data I would like to mark as thumb functions.

Behavior:

In the Linear Disassembly view, pressing "P" to mark new functions (on data that hasn't been identified as code) creates non-thumb ARM functions. In the right-click context menu, the "Create Function Here      P" option does not have a submenu that would allow creating a thumb function.

In the Hex Editor view, things are different. The default "P" action does create thumb functions, and the right-click context menu has a submenu that would allow me to pick ARM, Thumb, or whatever else I wanted.

Issue:

The default behavior of the "P" hotkey in Linear Disassembly view isn't what I wanted here. I'm not sure if it's incorrectly using ARM when it should be using thumb, or some heuristics about when to switch between ARM/thumb are just off in this case. Either way, I should have the option to override the default.

Hex Editor view behaves as desired (but obviously, is not convenient to access).

[percontation]

Quick illustration of the behavior difference, in case I described it poorly: https://gfycat.com/SpottedNegligibleBlacklemur

Also, noticed this comment: #883 (comment) , but as described above this seems to only be true in Hex Editor view and not Linear Disassembly view.

[bpotchik]

To create a Thumb function you can hit "P" on an odd address.

[psifertex]

The P behavior is as designed. As Brian mentioned, the arch is determined based on the address in the case of arm/thumb since that's how the hardware works. But once a default is set, we don't ask again since it's generally the right user experience.

The hex view not accounting for the odd/even doesn't surprise me since we don't have a binary view available there to use to check the address mapping (even if with a flat file the mapping is the same).

The issue about the right click create function being missing in linear view though is not something I can reproduce and would be a problem. Do you have a screen cap of that?

[percontation]

Ah. Odd address is very easy to remember :)

The issue about right-click create wasn't that it was missing; just that it was different from Hex View and does not give a submenu to select architectures (which made me think I had no way to create a thumb function from linear view, which was my main issue).

Linear view, no submenu to select arch:

Hex view has the submenu:

[percontation]

(I retract my suggestion that the linear view needs that arch-choosing submenu, now that I know how to make a thumb function there. Maybe it'd be good for consistency, though? idk.)

[psifertex]

Ahh, gotcha. Yeah, I see what you mean. Yeah, that's primarily a case of optimizing for the common use case. The situations where you want to make a function with the default architecture requires less clicks/navigation versus the rare situation where you can use a different view. That said, I could go either way on this one. We've got p for the fast use-case, having the linear sweep right-click might match hex might be fine. We'll talk it over internally some. In the meantime i'm going to close this one out since I think the biggest issue was knowing the arm/thumb addressing difference which fixes the primary problem for you.

[psifertex]

We are re-opening this as we believe the current behavior is indeed a bug and we wish to fix it. Here's what we believe should be done:

• Hitting "p" should always create a function of the default platform.
• We also believe that the associated platforms (arm/thumb in this case) should be sorted in the "Make Function at this Address" up top