Stylish haskell doesn't work in haskell-mode any more

[Mikolaj]

Others can't reproduce this, so it may be due to my emacs version (GNU Emacs 23.3.1, Ubuntu 12.04.1 LTS) or the cruft in my emacs configuration:

https://github.com/Mikolaj/mydotflies/blob/master/.emacs

Perhaps somebody else has a similar problem?

I can run stylish-haskell all right from command line, but it does not seem to have any effect anymore from haskell-mode. The commit that causes the behaviour is the last of the series that fixed many problems of stylish-haskell in haskell-mode

11b531c

With this commit, it apparently saves some extra stuff when I change and save the file:

Saving file /mikolaj/r/LambdaHack/Game/LambdaHack/Kind.hs...
Wrote /mikolaj/r/LambdaHack/Game/LambdaHack/Kind.hs [2 times]
Tags generated.

but the file is not affected apart of my original change.
For comparison, when I set

'(haskell-stylish-on-save nil)

I'm getting

Saving file /mikolaj/r/LambdaHack/Game/LambdaHack/Kind.hs...
Wrote /mikolaj/r/LambdaHack/Game/LambdaHack/Kind.hs
(No changes need to be saved)
Tags generated.

[Mikolaj]

It does not seem to be the fault of my emacs configuration, since it happens also with the following .emacs:

(custom-set-variables
 '(haskell-indentation-where-post-offset 1)
 '(haskell-interactive-mode-hide-multi-line-errors nil)
 '(haskell-process-prompt-restart-on-cabal-change nil)
 '(haskell-process-suggest-remove-import-lines nil)
 '(haskell-stylish-on-save t)
 '(haskell-tags-on-save t))
(load "~/r/haskell-mode/haskell-site-file")
(require 'cl)

[gregorycollins]

I can confirm this bug now -- my other laptop has Emacs 23.3.1 and exhibits the bad behaviour. You can see the error (which gets swallowed by the save hook) by running M-x haskell-mode-stylish-buffer.

[gregorycollins]

Bah, the call-process feature we're using was introduced in emacs 24. I'll fix :-(

[Mikolaj]

Bah, I've just upgraded to a backport of emacs 24. :D

[nrolland]

there is still a weird issue, as when haskell-stylish-on-save is enabled, the TAGS file supposed to be coming from haskell-tags-on-save is not actually generated

[ivan-m]

@nrolland maybe open a new issue for this as this one is labelled as being closed?

[Mikolaj]

there is still a weird issue, as when haskell-stylish-on-save is enabled, the TAGS file supposed to be coming from haskell-tags-on-save is not actually generated

I can't reproduce that with GNU Emacs 24.3.1 and haskell-mode from 2013-11-09 21:44:26 (doh, am I lazy with updating!).

[gracjan]

@nrolland: Thanks for reporting. We will need a reproducible case to debug this.

Also a new open issue would be a good idea so that new information is not lost.

[nrolland]

@gracjan I ma not sure what is the best way to have a reproducible case.

@Mikolaj : are you sure the file is actually generated ? I have the emacs display that tags have been generated but no file appears.
Very oddly enough, when calling directly M-x haskell-process-generate-tags, the file does get generated. FWIW, I tried to trace the calls down to haskell-process.el : (defun haskell-process-send-string (process string) but I dont see any difference in the string being sent when triggered on save VS manually invoked, which is puzzling...

":!cd /Users/nrolland/notes/utrecht/lab22/ && find . -name '.hs' -print0 -or -name '.lhs' -print0 -or -name '*.hsc' -print0 | xargs -0 hasktags -e -x
"

":!cd /Users/nrolland/notes/utrecht/lab22/ && find . -name '.hs' -print0 -or -name '.lhs' -print0 -or -name '*.hsc' -print0 | xargs -0 hasktags -e -x
"

invoking that command from cabal real does generate the tags.

[Mikolaj]

@nrolland: yes, the TAGS are automatically generated for me (but note I no longer have access to the old configuration from the time I reported the bug and with the new setup I see no erroneous behariour). Cheers!