Description
I noticed this by using company-mode
with the company-ghc
backend. It's an easy way to reproduce this problem.
I have a hook to enable company-mode
inside haskell-interactive-mode
. It works great usually, but recently I've noticed that my entire Emacs sometimes freezes up. After some investigation, I noticed that this happens only when I'm inside an IO action and either waiting (company-mode
by default triggers after a certain amount of time) or triggering the completion manually. A simple getLine
is enough, but a forever $ putStrLn ...
loop also works.
I don't think company-mode
nor company-ghc
are to blame here, since they seem to simply use haskell-process-get-repl-completions
. I killed my GHC process and had a look at the log:
-> getLine
<- ghc: signal: 15
-> :complete repl "\nasd"
<- 0 0 "\n"
<- ghc: signal: 15
ghc: signal: 15
<- Leaving GHCi.
Leaving GHCi.
Event: "finished
"
Process reset.
(everything after ghc: signal 15
happens when I call kill <ghc-process-id>
twice)
I noticed that this :complete repl ...
call always occurs before REPL+Emacs start hanging. So I looked in my .emacs.d
for files which send such a command, and haskell-process.el
in haskell-mode
showed up, specifically the haskell-process-get-repl-completions
. And sure enough, when I insert a (debug)
statement at the top of the function, I can see that this function is called when my problem would occur, and Emacs doesn't start hanging since it's not executing the rest.
This is possibly in some way related to #871. However, I'm not in multiline mode. To make sure, I explicitly executed :unset +m
before doing the same thing, and the problem still occurred.
Further info:
stack ghc -- --version
:The Glorious Glasgow Haskell Compilation System, version 7.10.3
- haskell-mode 20160518.945
M-x version
:GNU Emacs 24.5.1 (x86_64-apple-darwin13.4.0, NS apple-appkit-1265.21) of 2015-04-10 on builder10-9.porkrind.org
M-x eval-expression haskell-process-type
:stack-ghci
stack --version
:Version 1.0.4 x86_64