idris: install using haskell-stack #1630

ilovezfs · 2016-06-01T17:14:03Z

use haskell-stack to build idris to work around GHC 8 incompatibility
minor style changes to the test-do block

MikeMcQuaid · 2016-06-01T17:49:28Z

👍

ilovezfs · 2016-06-01T17:53:27Z

@MikeMcQuaid That would be lovely, but I expect it to fail. I want to see it fail on CI not just locally, to rule out any local issues.

However, I'm experimenting locally with a stack-based workaround.

ilovezfs · 2016-06-01T17:56:17Z

Yup, same failure as I get locally:

Building libraries...
/Applications/Xcode.app/Contents/Developer/usr/bin/make -C prelude build
../../dist/dist-sandbox-a2ddac0/build/idris/idris --build prelude.ipkg
./Builtins.idr:6:1: error: expected: declaration,
    end of input
||| The canonical single-element type, also known as the trivially 
^                                                                  
Type checking ./Builtins.idr
make[1]: *** [build] Error 1
make: *** [build] Error 2
cabal: Leaving directory '.'
cabal: Error: some packages failed to install:
idris-0.11.2 failed during the building phase. The exception was:
ExitFailure 2

ilovezfs · 2016-06-01T18:00:00Z

Stack-based workaround installs and passes the test:

==> Summary
🍺  /usr/local/Cellar/idris/0.11.2_2: 2,324 files, 936.0M, built in 8 minutes 39 seconds
iMac-TMP:Idris-dev joe$ brew test -v idris
Testing idris
==> Using the sandbox
/usr/bin/sandbox-exec -f /tmp/homebrew20160601-56408-1fj3ukz.sb /System/Library/Frameworks/Ruby.framework/Versions/2.0/usr/bin/ruby -W0 -I /usr/local/Library/Homebrew -- /usr/local/Library/Homebrew/test.rb /usr/local/Library/Taps/homebrew/homebrew-core/Formula/idris.rb -v
==> /usr/local/Cellar/idris/0.11.2_2/bin/idris /tmp/idris-test-20160601-56409-1w546mh/hello.idr -o /tmp/idris-test-20160601-56409-1w546mh/hello
==> /tmp/idris-test-20160601-56409-1w546mh/hello
iMac-TMP:Idris-dev joe$

ilovezfs · 2016-06-01T19:22:21Z

@MikeMcQuaid I've updated the PR with the haskell-stack based workaround.

Also, CC @zmwangx and @borsboom (RE: commercialhaskell/stack#848)

zmwangx · 2016-06-01T19:32:02Z

Formula/idris.rb

+      mkdir_p bin
+      system "stack", "install", "--local-bin-path=#{bin}"
+
+      install_dir = libexec/"stack/.stack-work/install/x86_64-osx/lts-5.11/7.10.3"


This is why we need stack install --prefix...

@MikeMcQuaid

install_dir = IO.popen("stack path --local-install-root") { |io| io.gets.chomp }

or

install_dir = `stack path --local-install-root`.chomp

?

install_dir = Utils.popen("stack path --local-install-root").chomp

borsboom · 2016-06-01T19:38:37Z

You might want to use the output of stack path --local-install-root instead of hard-coding .stack-work/install/x86_64-osx/lts-5.11/7.10.3 which could change in the future.

It looks like you're copying out of the .stack-work directory as a workaround for the lack of a stack --prefix option. I don't know if this is a problem with Idris in particular, but I'll just warn you that some Haskell packages are not relocatable because they end up with compiled-in hard-coded paths pointing to the original build directory (finding an elegant solution to this is the main holdup of supporting a stack --prefix option, but we recently got some code contributed that attempts to make these relocatable so hopefully we'll have a solution soon).

ilovezfs · 2016-06-01T19:46:59Z

You might want to use the output of stack path --local-install-root instead of hard-coding .stack-work/install/x86_64-osx/lts-5.11/7.10.3 which could change in the future.

Good idea.

for the lack of a stack --prefix option

Pretty please, with sugar on top.

I don't know if this is a problem with Idris in particular

It is, which is why I had to leave the .stack-work intact for everything except the docs and executables.

we recently got some code contributed that attempts to make these relocatable.

Nice! Do you happen to have a link?

ilovezfs · 2016-06-01T19:56:54Z

Woo hoo 🍏

MikeMcQuaid · 2016-06-01T19:57:45Z

Formula/idris.rb

+      system "stack", "setup"
+      system "stack", "build"
+
+      mkdir_p bin


Tiniest nit that normally we go for bin.mkpath

MikeMcQuaid · 2016-06-01T19:58:45Z

Tiny nits otherwise 👍

ilovezfs · 2016-06-01T19:59:58Z

@MikeMcQuaid Sweet.

ilovezfs · 2016-06-01T20:20:27Z

@MikeMcQuaid Also, we can optionally make the .stack-work visible using --work-dir=whatever_subfolder_name_you_want. Any preference between the default .stack-work or something else?

- use haskell-stack to build idris to work around GHC 8 incompatibility - minor style changes to the test-do block Issue Homebrew#1486.

ilovezfs · 2016-06-02T06:15:09Z

@DomT4 Any changes desired?

ilovezfs · 2016-06-02T09:26:16Z

@MikeMcQuaid any thoughts on #1630 (comment)? I don't know if there's any precedent for a significant invisible dot directory that's actually installed within the Cellar, but I don't mind the .stack-work default if you don't.

MikeMcQuaid · 2016-06-02T09:31:09Z

@ilovezfs I think I'd rather it's not in the Cellar if possible? I'm not sure I understand the contents or purpose, though, so I'm lacking context currently.

ilovezfs · 2016-06-02T09:40:24Z

It's required that it be installed since it contains the installation directory ("work-dir" is somewhat of a misnomer, or at least an incomplete description). The core issue is that some paths can get hard coded (which they indeed do in the case of Idris), and the stack work-dir must be a subfolder of the directory in which you're building. It can have whatever name you'd like, but the default is libexec/"stack/.stack-work" since we're building in libexec/"stack". Here's what's ultimately installed: https://gist.github.com/ilovezfs/0dd1839e08440f29471ff6aa40554762

MikeMcQuaid · 2016-06-02T09:43:13Z

I see, yuck. Normally I'm pretty 👎 on installation happening in the prefix but if it's unavoidable to fix up this package that's otherwise broken then a begruding 👍. I think in this case it's better to be a hidden/dotted work directory then, yeh.

ilovezfs · 2016-06-02T09:46:10Z

I see, yuck. Normally I'm pretty 👎 on installation happening in the prefix

@MikeMcQuaid hehe. Yes indeed. That's why I flagged what's happening here with a neon sign for you!

ilovezfs · 2016-06-02T09:52:52Z

On the bright side, this is a proof-of-concept for using stack with brew, even with its current limitations, and there's reason to be hopeful that the situation will improve in the future as there are upstream stack plans for the ability to locate the work-dir outside of the build path and, even more significantly, to add a --prefix option for installation.

ilovezfs · 2016-06-02T10:00:42Z

🚢 'd

BrewTestBot added the in progress Stale bot should stay away label Jun 1, 2016

ilovezfs force-pushed the idris-ci branch from 349c7df to 195e2b8 Compare June 1, 2016 17:32

ilovezfs changed the title ~~idris: trigger CI~~ idris: install using haskell-stack Jun 1, 2016

ilovezfs force-pushed the idris-ci branch from 195e2b8 to 3169dba Compare June 1, 2016 19:18

zmwangx reviewed Jun 1, 2016
View reviewed changes

borsboom mentioned this pull request Jun 1, 2016

Add stack build --prefix option commercialhaskell/stack#848

Open

MikeMcQuaid reviewed Jun 1, 2016
View reviewed changes

Formula/idris.rb

system "stack", "setup"

system "stack", "build"

mkdir_p bin

Copy link

Member

MikeMcQuaid Jun 1, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Tiniest nit that normally we go for bin.mkpath

idris: install using haskell-stack

53b81f4

- use haskell-stack to build idris to work around GHC 8 incompatibility - minor style changes to the test-do block Issue Homebrew#1486.

ilovezfs force-pushed the idris-ci branch from 3169dba to 53b81f4 Compare June 2, 2016 05:59

ilovezfs closed this in a93d454 Jun 2, 2016

BrewTestBot removed the in progress Stale bot should stay away label Jun 2, 2016

ilovezfs mentioned this pull request Jun 2, 2016

GHC 8 build fails with "error: expected: declaration, end of input" idris-lang/Idris-dev#3193

Closed

ilovezfs deleted the idris-ci branch June 3, 2016 10:21

Homebrew locked and limited conversation to collaborators May 4, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

idris: install using haskell-stack #1630

idris: install using haskell-stack #1630

ilovezfs commented Jun 1, 2016 •

edited

Loading

MikeMcQuaid commented Jun 1, 2016

ilovezfs commented Jun 1, 2016 •

edited

Loading

ilovezfs commented Jun 1, 2016 •

edited

Loading

ilovezfs commented Jun 1, 2016

ilovezfs commented Jun 1, 2016

zmwangx Jun 1, 2016

ilovezfs Jun 1, 2016

ilovezfs Jun 1, 2016

MikeMcQuaid Jun 1, 2016

borsboom commented Jun 1, 2016 •

edited

Loading

ilovezfs commented Jun 1, 2016

ilovezfs commented Jun 1, 2016

MikeMcQuaid Jun 1, 2016

MikeMcQuaid commented Jun 1, 2016

ilovezfs commented Jun 1, 2016

ilovezfs commented Jun 1, 2016

ilovezfs commented Jun 2, 2016

ilovezfs commented Jun 2, 2016

MikeMcQuaid commented Jun 2, 2016

ilovezfs commented Jun 2, 2016

MikeMcQuaid commented Jun 2, 2016

ilovezfs commented Jun 2, 2016

ilovezfs commented Jun 2, 2016 •

edited

Loading

ilovezfs commented Jun 2, 2016

idris: install using haskell-stack #1630

idris: install using haskell-stack #1630

Conversation

ilovezfs commented Jun 1, 2016 • edited Loading

MikeMcQuaid commented Jun 1, 2016

ilovezfs commented Jun 1, 2016 • edited Loading

ilovezfs commented Jun 1, 2016 • edited Loading

ilovezfs commented Jun 1, 2016

ilovezfs commented Jun 1, 2016

zmwangx Jun 1, 2016

Choose a reason for hiding this comment

ilovezfs Jun 1, 2016

Choose a reason for hiding this comment

ilovezfs Jun 1, 2016

Choose a reason for hiding this comment

MikeMcQuaid Jun 1, 2016

Choose a reason for hiding this comment

borsboom commented Jun 1, 2016 • edited Loading

ilovezfs commented Jun 1, 2016

ilovezfs commented Jun 1, 2016

MikeMcQuaid Jun 1, 2016

Choose a reason for hiding this comment

MikeMcQuaid commented Jun 1, 2016

ilovezfs commented Jun 1, 2016

ilovezfs commented Jun 1, 2016

ilovezfs commented Jun 2, 2016

ilovezfs commented Jun 2, 2016

MikeMcQuaid commented Jun 2, 2016

ilovezfs commented Jun 2, 2016

MikeMcQuaid commented Jun 2, 2016

ilovezfs commented Jun 2, 2016

ilovezfs commented Jun 2, 2016 • edited Loading

ilovezfs commented Jun 2, 2016

ilovezfs commented Jun 1, 2016 •

edited

Loading

ilovezfs commented Jun 1, 2016 •

edited

Loading

ilovezfs commented Jun 1, 2016 •

edited

Loading

borsboom commented Jun 1, 2016 •

edited

Loading

ilovezfs commented Jun 2, 2016 •

edited

Loading