gist-region results in `menu-bar-update-yank-menu: Wrong type argument: stringp, nil`

[jwiegley]

I just upgraded gh.el, and now when I attempt to gist a region I get:

menu-bar-update-yank-menu: Wrong type argument: stringp, nil

It turns out that the object passed to gist-created-callback is returning nil when (oref gist :html-url) is called.

[sigma]

that seems to indicate a malformed gist object, since the html-url field is never optional in the github api responses. Any chance you can provide a dump of the gist object as given to gist-created-callback ? Preferably a public gist, so that I can try to make sense of whatever discrepancy there might be.

[jwiegley]

Sure, how do I dump such an object?

[sigma]

just edebug the gist-created-callback function, and that should dump the object in the *edebug-trace* buffer

[jwiegley]

OK, I had to enable edebug tracing. The output is:

{ gist-created-callback args: ([eieio-class-tag--gh-gist-gist nil nil nil nil nil nil nil nil nil nil nil
   [eieio-class-tag--gh-user unbound unbound unbound unbound unbound] nil nil])

[jwiegley]

That seems like a pretty empty object; it makes me wonder why the callback was even invoked.

[sigma]

yeah that's odd. could you run the following to see if you get something similar?

(oref (gh-gist-get (make-instance 'gh-gist-api) "4b951b73220fb355887f6fd404f756b5") :data)

I'm also going to add some more type-checking in gh.el. Hopefully that should make such issues generate errors earlier

[jwiegley]

For that command I get back a fully elaborated object:

[eieio-class-tag--gh-gist-gist "4b951b73220fb355887f6fd404f756b5" "https://api.github.com/gists/4b951b73220fb355887f6fd404f756b5" "https://gist.github.com/4b951b73220fb355887f6fd404f756b5"
                               ([eieio-class-tag--gh-gist-gist-file "*scratch*.el" 4 "https://gist.githubusercontent.com/sigma/4b951b73220fb355887f6fd404f756b5/raw/b6601b5f799d6851b5501181abd2f58b3c77259b/*scratch*.el" "plop"])
                               nil "" "2016-11-12T01:48:52Z" "2016-11-12T06:39:26Z" "https://gist.github.com/4b951b73220fb355887f6fd404f756b5.git" "https://gist.github.com/4b951b73220fb355887f6fd404f756b5.git" 1
                               [eieio-class-tag--gh-user 16090 "https://api.github.com/users/sigma" "https://github.com/sigma" "sigma" nil]
                               ([eieio-class-tag--gh-gist-history-entry
                                 [eieio-class-tag--gh-user 16090 "https://api.github.com/users/sigma" "https://github.com/sigma" "sigma" nil]
                                 "8bc658b86fd8d81b0f27063d3a693e3c36102464" "2016-11-12T01:48:52Z"
                                 [eieio-class-tag--gh-gist-history-change 1 1 0]
                                 "https://api.github.com/gists/4b951b73220fb355887f6fd404f756b5/8bc658b86fd8d81b0f27063d3a693e3c36102464"])
                               nil]

[jwiegley]

@sigma Further thoughts? I'm really missing effortless gist integration, powered by gh.el. :) Sending pastes to co-workers takes much longer now.

[sigma]

so, the thing is I'm unable to reproduce this issue (on GNU Emacs 25.1.1). Which I suppose is good news cause there should be a way to make this work :)

From there, there are a few things that might help (and shouldn't hurt anyway):

• uninstalling/reinstalling gist/gh/pcache/marshal. This might help if for whatever reason there are inconsistencies between code compiled with various versions of EIEIO (which happens awfully frequently, as you probably know :))
• removing the ~/.emacs.d/var/pcache/gh directory (after exiting Emacs). This might help if somehow the cache got corrupted and was polluting the returned objects (I'm not sure I see how that'd be the case, but...)
• checking if you have any kind of proxy somewhere in the critical path? OK I'm just throwing random ideas at this point, I should stop :)

[jwiegley]

@sigma Try updating all the dependencies, and removing all pcache directories, but still get the same problem. I'm not using any proxies, since I'm trying this from home.

[sigma]

@jwiegley which version(s) of Emacs did you try that with?

[jwiegley]

I'm using 25.2pre, basically.

[sigma]

hmm that's really odd, I just built the tip of the emacs-25 branch (3f2b497cdddcf1f124f1c9f94354d5f457a4070f), started from an empty $HOME, installed gist.el from melpa, and still can't reproduce

[jwiegley]

@sigma Thanks for reminding me that I should have isolated it that way from the beginning. I'll narrow it down.

[jwiegley]

@sigma I've pared down to just this init.el, with no other settings, and I'm still getting the same problem, even after deleting all cache products relating to any of these libraries. Here's the entirety of my configuration now:

(add-to-list 'load-path "~/emacs/lisp/use-package")
(require 'use-package)
(use-package dash           :defer t :load-path "lib/dash-el")
(use-package gh             :defer t :load-path "lib/gh-el")
(use-package ht             :defer t :load-path "lib/ht-el")
(use-package logito         :defer t :load-path "lib/logito")
(use-package marshal        :defer t :load-path "lib/marshal-el")
(use-package pcache         :defer t :load-path "lib/pcache")
(use-package s              :defer t :load-path "lib/s-el")
(use-package tablist        :defer t :load-path "lib/tablist")
(use-package gist
  :load-path "site-lisp/gist"
  :bind ("C-c G" . gist-region-or-buffer))

I'm not loading any customization settings, or any other file besides use-package and these libraries, all of which I updated just now.

[jwiegley]

If you can't reproduce, I won't leave this open. I've restored my ability to gist by using a command-line gist utility that integrates nicely with Emacs.