thoughtpolice / cryptol-mode

A Cryptol major mode for Emacs.

The birds have vanished down the sky.

Now the last cloud drains away.

We sit together, the mountain and me,

until only the mountain remains.

Zazen on Ching-t’ing Mountain

cryptol-mode's Issues

Support type insertion for Cryptol 2

#14 only dealt with Cryptol 1, but Cryptol 2 is trickier to support as it doesn't yet support the -c command line option.

Literate file support.

Emacs 23 support.

We don't support Emacs 23 because it doesn't provide prog-mode. We should just derive from fundamental-mode instead I think.

I have an incoming patch for this; the ticket is just for tracking.

M-; prompts for comment syntax

M-; doesn't presently work the way I'd expect.

Expected behavior: I highlight a bit of text, and M-; comments it out

Observed behavior: I'm prompted for a comment syntax, after which M-; works.

Run check/exhaust on a given theorem.

Support for inserting inferred function types

Given a function definition, it'd be excellent if we could insert the functions inferred type signature much like haskell-mode and ghc-mod can.

This should be fairly simple in principle by invoking the compiler and just asking for the type (ghc-mod gets the benefit of being integrated via the GHC API;) for example (by default this will do bit interpretation, supporting all of Cryptol):

$ cat > test.cry
zip (a,b) = [| (x,y) || x <- a || y <- b |];
$ cryptol -qns test.cry -c ":type zip"
zip : {a b c d} ([a]b,[c]d) -> [min(a,c)](b,d)
$

Or:

$ cryptol -qns test.cry -c ":type dkatest"
dkatest : {ks bs n} (fin bs) => ({decrypt : ([ks],[bs]) -> [bs];
                                  encrypt : ([ks],[bs]) -> [bs]},
                                 [n]([ks],[bs],[bs])) -> [n]([bs],Bit)

Another question is what keybinding should be used. I am actually partial to M-t, which I picked up as a habit from ghc-mod. The original C-u C-c C-t used by haskell-mode could also work as well, but it's a bit longer. :)

Compile to Isabelle easily.

This might be complicated due to the fact I don't know how to get the Isabelle theories publicly; Galois must provide them.

Compile to VHDL easily.

cryptol-mode REPL contains unwanted control codes

When launching the REPL in cryptol-mode on a new file, I get the following errors:

                        _        _
   ___ _ __ _   _ _ __ | |_ ___ | |
  / __| '__| | | | '_ \| __/ _ \| |
 | (__| |  | |_| | |_) | || (_) | |
  \___|_|   \__, | .__/ \__\___/|_|
            |___/|_| version 2.1.0 (faab3ae)

�]0;cryptol��]0;cryptol�[error] can't find file: /Users/dchristiansen/tmp/Test.cry
cryptol> 
cryptol> 
Process cryptol finished
xo                        _        _
   ___ _ __ _   _ _ __ | |_ ___ | |
  / __| '__| | | | '_ \| __/ _ \| |
 | (__| |  | |_| | |_) | || (_) | |
  \___|_|   \__, | .__/ \__\___/|_|
            |___/|_| version 2.1.0 (faab3ae)

�]0;cryptol��]0;cryptol�Parse error at 3:1, unexpected end of input
cryptol> 

Github isn't showing the ^G characters that are also in it. Shouldn't these be stripped out?

Cryptol 2 support

Ensure we work on Windows.

Switching buffers means new files aren't loaded into the REPL.

Open A.cry and then do C-c C-l to load the REPL. Then open B.cry and do C-c C-l again - and only A.cry remains loaded in the REPL.

We should keep track of what file is currently loaded at the REPL and reload it when we switch buffers.

imenu only identifies theorems, not regular identifiers.

Compiler mode switching.

Includes things like C, SBV, VHDL, etc.

This is not directly related to the compiler tickets; this is only for quickly switching, since users might want to switch to refine theories in other backends.

This also might just be an easy stop gap to provide users now, and later we use it to switch mode+invoke compiler all at once easily.

Real indentation support.

REPL is broken on Linux

On my Ubuntu 12.10 machine, cryptol-mode is broken due to (I believe) incorrectly detected $TERM settings. When you launch the terminal, any keystrokes will cause cryptol to immediately exit. This is obviously a very bad bug.

Launching the REPL on my OS X machine Emacs correctly seems to set $TERM to dumb, and the REPL works just fine for playing around. It also does this for Linux, but it seems as if terminfo (which the cryptol executable is linked to) always reports that it cannot find the dumb mode despite having all the prerequisite packages installed.

I'll try to provide more details shortly but I realized I hadn't filed this.

Compile to C easily.

We should support both the C backend and translation via the (IMO much nicer) sbv backend.

Version/backend information is broken in Linux

Because of #1 when you use cryptol-version or cryptol-backends you get garbage output.

I have a fix for this. Creating just to keep track of it.