ELPA vs 9.0.1 release (still issues)

[Robert Horn <rjhorn@alum.mit.edu>]

I'm not enough of an ELPA guru to figure this out.  I tried org-mode
9.0.1 using ELPA, found some issues I'll describe, and find that using
the direct git version does not have those issues.

All of this is with Emacs 24.5.2, GTK+ version 2.24.31.  This is a fresh
setup, no other packages installed, but my .emacs is from a working
setup.

My first test of a new version is a babel invocation of R with a large
table sent as a variable to R:

#+name: glucose_analysis
#+begin_src R :file 1.png :results graphics :var data=glucose :width 1200 :height 600

library(ggplot2)
...

Test results:

1) The ELPA install from org repository of 20161118 had numerous compile
issues.  The immediately relevant one is:

Compiling file /home/hornrj/.emacs.d/elpa/org-20161118/org-element.el at Sun Nov 20 13:33:22 2016

In org-element--get-node-properties:
org-element.el:938:25:Warning: reference to free variable
    `org-planning-line-re'

In org-element--get-time-properties:
org-element.el:953:45:Warning: reference to free variable
    `org-planning-line-re'

In org-element--current-element:
org-element.el:3850:26:Warning: reference to free variable
    `org-planning-line-re'
org-element.el:3862:21:Warning: reference to free variable `org-clock-line-re'

In end of data:
org-element.el:6124:1:Warning: the following functions are not known to be defined: org-link-types,
    org-macro-extract-arguments


The invocation of R failed with an error that org-link-types was not
known.

2) I deleted all the .elc files in the ELPA directory

The result was an incredibly slow invocation of R.  It took 6 minutes to
parse and prepare the table.  (It's only about 2,000 line table.)  Then
it invoked R, generated the diagram, etc.  So it passed the test but was
incredibly slow.

3) I removed the ELPA installed directory, cloned the git repository,
did a "make compile" which generated no errors, and linked from elpa to
the lisp directory.

The test ran in a few seconds.

So there remains something wrong in the ELPA setup.

R Horn rjhorn@alum.mit.edu