Hopper Instruction Reference Tool Plugin.
Shamelessly ported from the excellent HopperRef plugin by zbuc.
Simply checkout or download the repository and use cmake/make to install it. Please also note that you need Xcode’s developer tools and sqlite3 to be available.
git clone https://github.com/0xc010d/InstructionReference.hopperTool.git cd InstructionReference.hopperTool git submodule init && git submodule update mkdir build && cd build cmake .. && make install
Restart Hopper Disassembler and you should see a “Show Current Instruction Reference” option in the “Tool Plugins” menu item.
Now for plagiarized documentation from nologic because he did all the hard work
Upon loading the script will look for SQlite databases in the same directory as the itself. The naming convention for the database files is [arch name].sql. The [arch name] will be presented to the user as choice.
The database has a table called ‘instructions’ and two columns called ‘mnem’ and ‘description’. The instructions are looked up case insensitive (upper case) by the mnem value. The text from description is displayed verbatim in the view.
To add support for more architectures simply create a new database with those columns and place it in the the script directory.
import sqlite3 as sq con = sq.connect("asm.sqlite") con.text_factory = str cur = con.cursor() cur.execute("CREATE TABLE IF NOT EXISTS instructions (platform TEXT, mnem TEXT, description TEXT)") con.commit()
When working with x86, I noticed that many instructions point to the same documentation. So, the plugin supports level referencing. Just place ‘-R:[new instruction]’ into description to redirect the loading. ‘new instruction’ is the target. So, when loading the script will detect the link and load the new target automatically.
cur.execute("INSERT INTO instructions VALUES (?, ?, ?)", ("x86", inst, "-R:%s" % first_inst))
Skeletons in the closet
The documentation database was created using a rather hackish screen scraping technique by the x86doc project which I forked. So, there are probably some strange characters or tags in the text. At least, it is a mechanical process so I expect that the information is correct relative to the original Intel PDF.