`isMinTTY` no longer detects recent `mintty`s

[RyanGlScott]

isMinTTY incorrectly returns false on recent versions of mintty.

Current Behavior/Steps to Reproduce

To reproduce, run the following in GHCi:

> import System.Win32.MinTTY
> isMinTTY
False
> :! mintty --version
mintty 3.5.1 (x86_64-pc-msys)
© 2013/2021 Andy Koppe / Thomas Wolff
License GPLv3+: GNU GPL version 3 or later
There is no warranty, to the extent permitted by law.

Your Environment

• Version used: I've tried this with multiple versions of Win32, including 2.10.0.0 and 2.12.0.1. I suspect that the mintty version is the real culprit, however.
• Operating System and version: Windows 10 Home, 10.0.19042 Build 19042

[RyanGlScott]

With some printf debugging:

diff --git a/System/Win32/MinTTY.hsc b/System/Win32/MinTTY.hsc
index 97218b0..ffdb749 100644
--- a/System/Win32/MinTTY.hsc
+++ b/System/Win32/MinTTY.hsc
@@ -79,6 +79,7 @@ isMinTTYHandle h = do
 isMinTTYVista :: HANDLE -> IO Bool
 isMinTTYVista h = do
     fn <- getFileNameByHandle h
+    putStrLn fn
     return $ cygwinMSYSCheck fn
   `catch` \(_ :: IOError) ->
     return False

I get:

> isMinTTY
\msys-dd50a72ab4668b33-pty0-to-master-nat
False

Here is how cygwinMSYSCheck is currently implemented:

https://github.com/haskell/win32/blob/cd69b1944f1afa4f85bca314e91c56f00cf3f921/System/Win32/MinTTY.hsc#L93-L98

This assumes that -master will be at the very end, but in this case, we have -master-nat. I wonder what the right thing to do is? Here is what the Rust atty library does:

    // This checks whether 'pty' exists in the file name, which indicates that
    // a pseudo-terminal is attached. To mitigate against false positives
    // (e.g., an actual file name that contains 'pty'), we also require that
    // either the strings 'msys-' or 'cygwin-' are in the file name as well.)
    let is_msys = name.contains("msys-") || name.contains("cygwin-");
    let is_pty = name.contains("-pty");
    is_msys && is_pty

git-for-windows does something similar.

Perhaps we should remove the -master check? I blindly copied that part from the MinTTY detection code from GHC at the time, but perhaps that part was never really correct to begin with.

[Mistuke]

Hi, thanks for looking into this,

Dropping the master check seems sensible to me. The one thing I'm wondering about is whether we still do the right thing on newer mintty. Specifically newer mintty have support for conpty[1] which is an actual pty without the quirks of mintty.

https://github.com/mintty/mintty/issues/804

So my question is if the behavior we're guarding against with these checks are still valid.

But I guess that's a different question..

[1] https://devblogs.microsoft.com/commandline/windows-command-line-introducing-the-windows-pseudo-console-conpty/

[RyanGlScott]

Interesting. What is the easiest way to enable ConPTY support in mintty so that I could test it?

[Mistuke]

You have to have an msys runtime based on cygwin 3.1.0 and up, so uname -r should be 3.1.0+, you can upgrade with pacman -Syuu

After upgrading the runtime you can upgrade your packages (repeat the same command) which should get you an updated mintty.

In particular with conpty support you should be able to use any terminal, like Microsoft Terminal with msys2 and it should all work the same as MinTTY.

[RyanGlScott]

As far as I can tell, I already have a sufficiently recent version of Cygwin:

$ uname -r
3.2.0-340.x86_64

If I launch cmd.exe from within MSYS2's mintty and repeat the same exercise, I get:

> isMinTTY
\msys-dd50a72ab4668b33-pty0-to-master-nat
False

[Mistuke]

As far as I can tell, I already have a sufficiently recent version of Cygwin:

$ uname -r
3.2.0-340.x86_64

If I launch cmd.exe from within MSYS2's mintty and repeat the same exercise, I get:

> isMinTTY
\msys-dd50a72ab4668b33-pty0-to-master-nat
False

Ah that's not what I meant. So currently these mintty detections are mostly being used to know when we need to behave differently as MinTTY was not a true tty. But when it's backed by conpty it should be.

Basically I was just wondering (out loud, not specifically to you) whether GHC(i) still needs to enable "mintty" mode

[RyanGlScott]

Even if MinTTY does support ConPTY, GHCi still behaves quite differently under MinTTY than it does other Windows consoles. For instance, GHCi's tab completion and Ctrl-C handling are still completely broken under MinTTY.