Compiling probekins Reloaded

I am so sloppy: matplotlib never installed.

When I looked at stlvis.py and saw it had other 'include' lines that it didn't complain about, I ran a find on one of them (numpy) and then one on matplotlib. Tracing my steps backward I got all the way back to easy_install where, again, I found I hadn't read "all the way to the end" and _this time_ saw.

I tried to update (easy_install had picked-the-lastest for matplotlib):
$ sudo easy_install numpy
Searching for numpy
Best match: numpy 1.3.0
Adding numpy 1.3.0 to easy-install.pth file

but it reports that it's happy with the current 1.3.0 version. Same with
$ sudo apt-get install python-numpy

My quest now: upgrade numpy to >1.5. That, and reading-comprehension lessons.

See this link posted by someone trying the same thing for the same reasons

www.black-glass.com/?p=306

I was well into a compile-from-source step (numpy 1.8.0) when I came back to read your link.

While all that text was flying by, I couldn't help thinking of each step as a "point of failure"... I can't recommend this step for anyone else--because I can't understand it...

When it finished successfully I couldn't find it , so I did a "find /usr " again and found numpy 1.8.0 in /usr/local/lib/ instead of /usr/lib (confirmed by cat'ing a version.py file I found in its dir). I had seen a reference to an environment variable ($PYTHON[_]PATH?) to cause python to find a particular version first, but I took the cruder approach of renaming /usr/lib/.../numpy/ to numpy-1.3.0 and added a link there over to /usr/local/lib/.../numpy-1.8.0 (which I'd also renamed from numpy).

I say "successfully" because
$ sudo easy_install matplotlib
proceeded without complaining about numpy.
Turns out, this fails:
$ sudo easy_install dateutil
Of course, I don't care about "date axis support," but I don't get an opportunity to tell it that.

Success!... of a sort.

$ sudo easy_install matplotlib 1.2.1 # FAIL, can't find
$ sudo easy_install matplotlib-1.2.1 # FAIL
$ sudo easy_install matplotlib_1.2.1 # FAIL

$ sudo easy_install "matplotlib-1.2.1" # FAIL BUT, "a clue"
error: Not a URL, existing file, or requirement spec: 'matplotlib 1.2.1'

$ sudo easy_install downloads.sourceforge.net/project/matplo...plotlib-1.2.1.tar.gz
successfully installed the earlier version, and stlvis.py was able to generate a plot.

$ man easy_install # FAIL

IMHO easy_install --help doesn't suggest this "direct" method for installation.

This kind of circular dependency hell is quite common when you try to put a new app on an old distro.
(but not restricted to that scenario)
Ubuntu 10.04 is old, it is out of long term support and they ceased building any new packages for it some while before that

The app requires lib1 of a certain version, you install it and find that lib1 requires lib2 and lib3 of certain versions, then lib3 requires lib4 of a certain version etc

If you are lucky and none of the later libs now conflict with other libs your app relies upon, you get it all installed and eventually your new app works.

Next day you find that half a dozen of your old apps have stopped working because the libs were not backwardly compatible

ArcEye, (Mind if I call you "Eye?" "YES!" okay...)

I've got to say that without the fellowship (of the <something, needs to rhyme with Ring>) that can be found here, these kinds of adventures would end in a lot more veins being opened.

Along the lines you observed, before I left this morning, after I switched back to numpy-1.3.0, the older matplotlib-1.2.1 said "numpy 1.3.0 is too old: need AT LEAST 1.4.0..."

The way I set it up, I can still "throw the switch" to point over to 1.8.0, but given how old 10.04 is, I think the wisest choice is to try to operate on the "least modern" files one can get away with, so I'm going to keep playing around to see how far back down the food chain I can go with matplotlib and still satisfy stlvis.py. (But this might have gotten beyond wanting results, to wanting revenge... ;^) )

Thank you for your truly helpful pointers. I'm thinking LinuxCNC has to migrate to a newer LTS sometime, but if it doesn't, I'm going to have to take more care not to upset 10.04.