example: Create sleigh-lift example program

[tetsuo-cpp]

Closes #1

• Build in CI
• Use more stringent compiler warnings
• Make this part of the sleigh CMake project instead of a separate one
• Figure out what's wrong with the remill-lift example
• Figure out whether the excessive Pcode is incorrect

[tetsuo-cpp]

This still needs to be cleaned up but I wanted to check that it's more or less what you were expecting. I've modelled it after remill-lift.

At the moment, it works like this:

tetsuo@Alexs-MacBook-Pro build % ./sleigh-lift disassemble ~/Build/install/share/sleigh/Processors/x86/data/languages/x86-64.sla 4881ecc00f0000
0x00000000: SUB RSP,0xfc0
tetsuo@Alexs-MacBook-Pro build % ./sleigh-lift pcode ~/Build/install/share/sleigh/Processors/x86/data/languages/x86-64.sla 4881ecc00f0000
(register,0x200,1) = INT_LESS (register,0x20,8) (const,0xfc0,8)
(register,0x20b,1) = INT_SBORROW (register,0x20,8) (const,0xfc0,8)
(register,0x20,8) = INT_SUB (register,0x20,8) (const,0xfc0,8)
(register,0x207,1) = INT_SLESS (register,0x20,8) (const,0x0,8)
(register,0x206,1) = INT_EQUAL (register,0x20,8) (const,0x0,8)
(unique,0x12c00,8) = INT_AND (register,0x20,8) (const,0xff,8)
(unique,0x12c80,1) = POPCOUNT (unique,0x12c00,8)
(unique,0x12d00,1) = INT_AND (unique,0x12c80,1) (const,0x1,1)
(register,0x202,1) = INT_EQUAL (unique,0x12d00,1) (const,0x0,1)

I don't know why there's so much Pcode for that instruction.

I noticed that the disassembly doesn't match for the Remill example here. It looks like this for me:

tetsuo@Alexs-MacBook-Pro build % ./sleigh-lift disassemble ~/Build/install/share/sleigh/Processors/x86/data/languages/x86-64.sla c704ba01
0x00000000: MOV word ptr [SI],0x1ba

I'll have to look into that also.

[pgoodman]

I don't know why there's so much Pcode for that instruction.

I think that's the p-code for SUB RSP,0xfc0? If so, then the complexity is due to x86 flags computations, in which case that p-code looks absolutely fine to me :-)

Overall this seems good. Can you make some convenience functions for automatically discovering the locations of these SLA files? Remill has the same problem, and I have some search paths hard-coded [1], some of which are CMake variables. Probably you could do a better job of what I did in Remill by using CMake's ability to configure a file, substituting @ variables.

[1] https://github.com/lifting-bits/remill/blob/master/lib/BC/Util.cpp#L497-L548

[pgoodman]

This looks good. A few superficial edits requested.

Otherwise, do you have a sense on how extensible the assembly printers are? Do they break the assembly down into any kind of structure, such as tokens or trees, or would I need to assemble then disassemble to get that?

[tetsuo-cpp]

Otherwise, do you have a sense on how extensible the assembly printers are? Do they break the assembly down into any kind of structure, such as tokens or trees, or would I need to assemble then disassemble to get that?

It doesn't seem to provide a tree structure. You just get an address, an instruction mnemonic and then a "body" which is basically just the arguments.

[tetsuo-cpp]

Can you make some convenience functions for automatically discovering the locations of these SLA files?

This isn't trivial so I'd prefer to make a follow up change with this (I've made #12 to track it). I'm guessing this should be exposed in the header that we discussed above since other projects will want this too.

The problem is that the .sla files aren't checked in. They're generated from .slaspec files being fed into the sleigh_opt compiler, and they're all generated in different directories alongside the .slaspec files. So I really need some platform independent way of copying all the .sla files into a single directory within the installation and then passing that into a CMake variable such that it can be used in the code.

I think the right thing to do is to stop using sleigh_opt in batch mode and instead have a foreach that loops over spec_file_list and invokes the compiler on each one. And use get_filename_component on the .slaspec to put together a desired filepath to write the .sla file to.

[pgoodman]

I think the right thing to do is to stop using sleigh_opt in batch mode and instead have a foreach that loops over spec_file_list and invokes the compiler on each one. And use get_filename_component on the .slaspec to put together a desired filepath to write the .sla file to.

Try to use std::filesystem::path when/where possible. We have some repos that have a FindFilesystem.cmake to make sure we always find it; check Dr. Lojekyll repo.

[tetsuo-cpp]

Try to use std::filesystem::path when/where possible. We have some repos that have a FindFilesystem.cmake to make sure we always find it; check Dr. Lojekyll repo.

Sounds good.

I was referring more to the CMake code that runs on installation though. At the moment the generated .sla files are littered throughout the installation so I need to fix this so they're all in one directory. Then I can point the C++ helper at the directory containing the .sla files and use the std::filesystem stuff to find them.

Though, this isn't really related to sleigh-lift so I'd prefer to just get this in and follow up with the header + helpers to find .sla files.

[pgoodman]

Lets pull out that one "megaheader" include thing with the pragma's and call it a PR :-)