warbo-utilities: 208a50dfe2270315655c5ba70ff98ec7f6e45068

     1: From: Chris Warburton
     2: Date: Mon, 12 Jun 2017 13:50:08 +0100
     3: State: resolved
     4: Subject: Make emacsclient work as artemis editor
     5: Message-Id: <afbc1cc47d3479cf-0-artemis@nixos>
     6: resolution: fixed
     7: 
     8: git-artemis makes files called /tmp/tmpXXX.txt for issues, opens them in
     9: EDITOR, then uses the result to create the actual issue. Fair
    10: enough. The problem is, we want to use emacsclient as EDITOR, and that
    11: will treat /tmp/tmpXXX.txt as a text file (and rightly so!).
    12: 
    13: What we need to do is open the file in emacsclient, then trigger some
    14: Emacs Lisp to set the mode. Note that we can't do this with a single
    15: emacsclient call, since opening files and evaluating code are treated as
    16: mutually exclusive options.
    17: 
    18: We must open the file first, so that the buffer's available when we run
    19: our ELisp. However, our ELisp evaluation will exit immediately, rather
    20: than waiting for us to finish editing the file.
    21: 
    22: So, the solution seems to be:
    23: 
    24:  - Write a wrapper around git-artemis. Might as well call it
    25:    "artemis". This will set the EDITOR to a custom script.
    26:  - The custom script accepts a filename as $1, then calls emacsclient
    27:    with that filename, using & to run it in the background.
    28:  - We sleep for a second, to give Emacs chance to open the file, then we
    29:    run emacsclient again, passing '--eval "(some-custom-lisp \"$1\")"'
    30:    where some-custom-lisp will switch to the given filename's buffer and
    31:    set the right mode.
    32:  - We then need to `wait` for the first emacsclient process to finish
    33:    (i.e. wait until we've finished editing). Otherwise we'll exit
    34:    immediately, artemis will see the unmodified text and abort (since it
    35:    checks whether the text is modified). Meanwhile we'll be in Emacs,
    36:    blissfully unaware that artemis has already finished, and the issue
    37:    we're entering will just end up cluttering /tmp without being
    38:    processed.