Is it possible to work interactively?

[Mercury13]

The main feature of IDA is: it doesn't have to be perfect. The user presses C, and the part turns to code.
You are probably the closest one to IDA among open-source.

[Dax89]

That feature along with instruction patching are the blockers of beta2 and the ones I'm working on.

Nightly builds already provides instruction patching (x86 only for now), but not the "C press" shortcut to turn data to code and reignite the analysis.

So yes, I can add what are you asking because it's just a different entry point of the same problem: in both cases engine needs to be triggered again to do a partial analysis because of a model change.

[Mercury13]

Thank you. When are you going to finosh it somehow?

[Dax89]

In my plans is expected to be released next week.

But it can be delayed if I find something wrong, as I said changing bytes is easy but the internal disassembler state needs to be resynced in order to stay consistent.

[Mercury13]

What I am exploring... A program written in rather archaic Delphi (7 or even earlier), unprotected.
All RAD tools tackle one problem: form files are data, and that data has to point to code somehow.
Usually by "stringly-typing".
Of course, exploration from entry point onwards won't find that code.
I found a key string, there is something looking like code nearby, but it does not disassemble.

[Mercury13]

My own program Unicodia has almost no stringly-typing, just because I connect signals to slots in the constructor, but as pointers to functions still go through data, I cannot make sure that your tool will disassemble them. But still, INTERACTIVE disassembler doesn't have to be perfect, instead it relies on operator's reasoning.

[Dax89]

What I am exploring... A program written in rather archaic Delphi (7 or even earlier), unprotected.

Have you tried DeDe (Delphi Decompiler)? Don't know if still available it was an old tool for Delphi programs.

REDasm doesn't ship a delphi decompiler, but it's planned: currently it's only capable to find and recover VB5/6 events which is not what you need.

Yes, in your case converting data to code it's the right approach without a language specific analysis

[Mercury13]

DeDe does not work.
I used to work in IDA, but for some reason it's forbidden.