RE: Multiple versions of Org in load-path problem

[Cook, Malcolm]

>Ihor Radchenko <mailto:yantar92@posteo.net> writes:
>
>> David Masterson <mailto:dsmasterson@gmail.com> writes:
>>
>>> I went so far as to add the following to early-init.el:
>>>
>>> (if (featurep 'org) (unload-feature 'org))
>>> (add-to-list 'load-path (expand-file-name "~/.emacs.d/elpa/org-9.6")
>>> (load-library "org")
>>>
>>> The load-library still fails with the org-assert-version error.
>>
>> Indeed. It is to be expected. M-x package-install should not err here
>> though. In theory.
>
>My testing says that, in early-init.el, Org is not yet a feature, so the
>built-in Org-9.3 is not yet loaded and the first 'if' fails. Adding the
>path for Org-9.6 should put it at the front of the load-path and
>(hopefully) override Org-9.3 if it is in the load-path (I think I
>checked that). The error in the load-library is, therefore, a problem
>with Org-9.6 installed by package-install. 
>
>> I suggest you to read `org-assert-version'. It aims to catch the
>> situation above and similar.
>
>This is chicken and egg problem. I can't use 'org-assert-version'
>unless I can get Org to load.
>
>> Note, however, that installation problem is different. Or rather it
>> should be different.
>>
>>> Could it be that, when you package-install Org-9.6, it runs through a
>>> check where (I think) it compiles the package? I get a lot of warnings
>>> and 34(?) failed files. Could it be that some of the files are not
>>> built and, so, when I load-library, I'm picking up some of the wrong
>>> files from the built-in Org-9.3? Do we need to remove all versions of
>>> Org that is not the current version from the load-path at the very
>>> beginning of org.el to ensure no fall-through? Kind of a hack, but...

Hi,

I've been reading along as I've been similarly hounded in years past with org 
version changes.

It really sounds like you compiled org-9.6 with Org-9.3 already loaded.

I'd recommend reinstalling.

Here is one way to do it that is pretty safe:

        > emacs -Q -batch -eval "(progn (require 'package) (package-initialize) 
 (package-refresh-contents) (package-install 'org))"

If after doing this, and running emacs, you find M-x org-version does not 
return 9.6, or you find the package assert problem, I'd try the more paranoid:
        > emacs -Q -batch -eval "(progn (require 'cl-seq) (delete (car 
(cl-member "lisp/org" load-path :test  #'string-match)) load-path) (require 
'package) (package-initialize)  (package-refresh-contents) (package-install 
'org))"

And possibly consider adding to your init.el at the very top:

        (require 'cl-seq) (delete (car (cl-member "lisp/org" load-path :test  
#'string-match)) load-path)

YMMV,

~ Malcolm

>>
>> Maybe. It should not. That's why I raised
>> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=59882
>> I'd like to see if Emacs devs have anything to say here.
>
>Good. I'm not used to bug reporting in Emacs, so I raised a question on
>use-package in mailto:emacs-devel@gnu.org, but they pointed me at your bug.
>I'll see if I can add anything there.
>
>-- 
>David Masterson
>