have stack ghci work more like cabal repl in terms of sandboxing .hi and .o files

[wolftune]

After running stack ghci, all my .hs files have corresponding .hi and .o files. I don't like having those pollute my project directory, so I'm hoping there could be a way to run this where those are stored in some separate hidden or sandboxed place…

[pharpend]

FWIW, I can reproduce this:

Edit well, I could reproduce it in my existing snowdrift tree, but not on a fresh one.

git clone git://github.com/snowdriftcoop/snowdrift.git
cd snowdrift
stack setup
stack build

Due to a GHC bug, this will probably fail a few times. Just keep running stack build until it succeeds.

Finally

stack test
stack ghci

I'm getting a segfault on stack test, which... I don't think I've ever seen Haskell code segfault before.

If you ever do get to 'stack ghci', there will be .hi and .o files littered throughout the source tree.

[DanBurton]

+1

On Monday, July 6, 2015, Peter Harpending [email protected] wrote:

FWIW, I can reproduce this

—
Reply to this email directly or view it on GitHub
#529 (comment)
.

-- Dan Burton

[chrisdone]

This is easy enough, will do.

[chrisdone]

Implemented in 712c965. @wolftune Confirm?

[wolftune]

@chrisdone how do I install this to test? Do I need to run stack upgrade --git ? I'm a little confused about the strange docs around upgrading with vague references to "this" in the upgrade stuff and then link to an issue comment…

[chrisdone]

stack upgrade --git works. I tend to just have the git repo open and do stack install.

[wolftune]

@chrisdone it appears that I didn't have stuff polluted with .o files but I still got .hi files.

I also had a separate problem that appears to be related to failing to import Prelude or something where I had a failure at one point even though I did end up at the prompt anyway…

[chrisdone]

Oh, I always thought -odir handled interface files too. Turns out there's a -hidir flag, which I've added now in fbcdf56.

[wolftune]

Yay, it's all working!

[lukexi]

Just FYI, there's also -outputdir which I believe redirects all output files in one go (off the top of my head there's also e.g. dyn_o files in dynamic mode, and stub files for foreign exports, so it might be an easier way to go)