Fix #18691 fix to #17572 fix to R.

[charpent]

#18691 is broken on fresh install because the sage-env script is invoked at a point where $SAGE_LOCAL or $DOT_SAGE do not exist yet.

[kcrisman]

Changing priority, just change back if I misunderstand the severity of the problem.

[jdemeyer]

Should this

if [ -d "$SAGE_LOCAL" ] ; then
    R_MAKEVARS_SITE="$SAGE_LOCAL/lib/R/share/Makevars.site" && export R_MAKEVARS_SITE

be

if [ -d "$SAGE_LOCAL/lib/R/share" ] ; then
    export R_MAKEVARS_SITE="$SAGE_LOCAL/lib/R/share/Makevars.site"

---

New commits:

​9a9c4cc	Original tarball dropped in place.
​769fc49	Restored spkg-src.
​b46fa1a	Restored ORIGINAL non-functional spkg-src as par Nathaann Cohen's request.
​85de560	Condition Makevars file creation to existence of the relevant trees.

[charpent]

Jeroen : I'm afraid you mixed this ticket with #18820. I do not understand what the hell 9a9c4cc, 769fc49 and b46fa1a, which belong to #18820, are doing in the present #18835.

Please tell me how to undo this mess...

[jdemeyer]

Replying to charpent:

Jeroen : I'm afraid you mixed this ticket with #18820.

I didn't do anything, I just made a comment...

Please tell me how to undo this mess...

Just cherry-pick the relevant commit on top of #18691.

[jhpalmieri]

I lowered the priority. The build is not broken, but a needless error message is (prominently) displayed when building from scratch.

[jhpalmieri]

I know this is not marked as needing review yet, but the proposed change

if [ -d "$SAGE_LOCAL" ] ; then

is not good enough: you need to check whether the directory $SAGE_LOCAL/lib/R/share exists before doing echo foo > "$R_MAKEVARS_SITE".

[charpent]

I managed to hose my branch. I still have to learn how to revert certain commits.

Stay tuned (but not until tomorrow or late tonight). And accept my apologies.

Emmanuel Charpentier

[kcrisman]

I lowered the priority. The build is not broken, but a needless error message is (prominently) displayed when building from scratch.

Got it! Thanks for clarifying.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​94275f8	Revert "Restored ORIGINAL non-functional spkg-src as par Nathaann Cohen's request."
​652a004	Revert "Restored spkg-src."
​cc32c78	Revert "Original tarball dropped in place."

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​0bbcb05

Condition creating Sage-specific R files to the existence of their destination trees.

[charpent]

Reading the git doc was ... instructive ...

This one seems good (builds from make distclean without error and passes all tests of make ptestlong ==> needs_review.

[jdemeyer]

I think git might get confused when merging this...

Please just remove the bad commits instead of reverting them.

[charpent]

Replying to jdemeyer:

I think git might get confused when merging this...

Please just remove the bad commits instead of reverting them.

How do you do that ?

[jdemeyer]

There are many possibilities, I guess

git rebase -i

would be the easiest. Then you can edit, squash, remove, reorder commits...

Of course, this qualifies as "rewriting history" which is something you normally should not do for a ticket under review. However, in this case, it is the best solution.

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​18a1262

Condition Makevars file creation to existence of the relevant trees.

[charpent]

Done. I learned something useful. needs_review, again.

However, something worries me. I've read again and again and again (in the Developer's guide, on sage-devel threads, on various ticket brawls...err...discussions) thar "rewriting history" was, a priori a bad idea. I suppose that, in the present case, it is admissible because nobody has yet merged this contribution in anything else. Is this correct ?

Update : this, on top of 6.8beta6, builds with no errors and passes testlong successfully

[charpent]

Update : on top of 6.8beta7, this gives one ptestlong failure :

----------------------------------------------------------------------
sage -t --long --warn-long 67.6 src/sage/structure/dynamic_class.py  # 2 doctests failed
----------------------------------------------------------------------

However, this doctest passes standalone :

charpent@asus16-ec:/usr/local/sage-6.8$ sage -t --long --warn-long 67.6 src/sage/structure/dynamic_class.py
Running doctests with ID 2015-07-03-19-21-20-3619480c.
Git branch: mabranche
Using --optional=mpir,python2,sage,scons
Doctesting 1 file.
sage -t --long --warn-long 67.6 src/sage/structure/dynamic_class.py
    [69 tests, 0.23 s]
----------------------------------------------------------------------
All tests passed!
----------------------------------------------------------------------
Total time for all tests: 0.3 seconds
    cpu time: 0.2 seconds
    cumulative wall time: 0.2 seconds

So I think it's a glitch possibly due to high load.

Still needs_review

[jhpalmieri]

I get the same doctest failure in plain 6.8.beta7. I don't think it's related to this ticket.

[jhpalmieri]

This looks good to me now.