sagenb-0.8.p1.spkg: broken dependencies, requires internet conn. to build

[sagetrac-GeorgSWeber]

This was first noted by John Cremona.
On a computer that is not connected to the internet, the build fails:

Processing sagenb-0.8-py2.6.egg
removing '/home/john/sage-4.4.1.rc0/local/lib/python2.6/site-packages/sagenb-0.8-py2.6.egg'
(and everything under it)
creating /home/john/sage-4.4.1.rc0/local/lib/python2.6/site-packages/sagenb-0.8-py2.6.egg
Extracting sagenb-0.8-py2.6.egg to
/home/john/sage-4.4.1.rc0/local/lib/python2.6/site-packages
sagenb 0.8 is already the active version in easy-install.pth
Installing jmol script to /home/john/sage-4.4.1.rc0/local/bin
Installing sage3d script to /home/john/sage-4.4.1.rc0/local/bin

Installed /home/john/sage-4.4.1.rc0/local/lib/python2.6/site-packages/sagenb-0.8-py2.6.egg
Processing dependencies for sagenb==0.8
Searching for mechanize<0.2dev
Reading http://pypi.python.org/simple/mechanize/
Download error: [Errno -2] Name or service not known -- Some packages
may not be found!
Reading http://pypi.python.org/simple/mechanize/
...

On a computer with internet connection, one finds in the install.log:

Processing dependencies for zope.testbrowser==3.8.1
Searching for mechanize<0.2dev
Reading http://pypi.python.org/simple/mechanize/
Reading http://wwwsearch.sourceforge.net/mechanize/
Best match: mechanize 0.1.11
Downloading http://wwwsearch.sourceforge.net/mechanize/src/mechanize-0.1.11.tar.gz
Processing mechanize-0.1.11.tar.gz
Running mechanize-0.1.11/setup.py -q bdist_egg --dist-dir /tmp/easy_install-XXPn64/mechanize-0.1.11/egg-dist-tmp-fxjRAP
no previously-included directories found matching 'docs-in-progress'
Removing mechanize 0.2.0 from easy-install.pth file
Adding mechanize 0.1.11 to easy-install.pth file

...

So the solution seems to be pretty easy:

Make a new sagenb-0.8.p2.spkg, by throwing out "mechanize-0.2.0.tar.gz" under /src/, adding back the old "mechanize-0.1.11.zip", and finally adjusting the corresponding line 6 in the spkg-install file.

Apply this spkg:

• http://sage.math.washington.edu/home/mvngu/spkg/standard/sagenb/sagenb-0.8.p2.spkg

CC: @williamstein @TimDumol

Component: build

Author: Tim Dumol

Reviewer: Minh Van Nguyen, Georg S. Weber

Merged: sage-4.4.2.alpha0

Issue created by migration from https://trac.sagemath.org/ticket/8861

[sagetrac-GeorgSWeber]

comment:1

(corrected numbering of the package cited)

[sagetrac-GeorgSWeber]

comment:2

I just couldn't resist and did exactly that, the spkg is here:

http://sage.math.washington.edu/home/weberg/spkg/sagenb-0.8.p2.spkg

I note this still as "needs work" for three reasons:
1.
My own test is not done yet --- one needs to do it on a computer without internet connection (I'll do overnight and report tomorrow).

The sagenb spkg inexplicably misses the required hg repository (to able to monitor changes in spkg-install for example).

In the file SPKG.txt, adding a note about testing on a computer without internet connection seems appropriate.

[TimDumol]

comment:3

Thanks for the quick work!

Replying to @sagetrac-GeorgSWeber:

I just couldn't resist and did exactly that, the spkg is here:

http://sage.math.washington.edu/home/weberg/spkg/sagenb-0.8.p2.spkg

I note this still as "needs work" for three reasons:
1.
My own test is not done yet --- one needs to do it on a computer without internet connection (I'll do overnight and report tomorrow).

The sagenb spkg inexplicably misses the required hg repository (to able to monitor changes in spkg-install for example).

The sagenb spkg is automatically generated by src/sagenb/spkg-dist, therefore the hg repository is actually in src/sagenb.

In the file SPKG.txt, adding a note about testing on a computer without internet connection seems appropriate.

Done.

I've posted the version with the changes in the hg repo here: http://sage.math.washington.edu/home/timdumol/sagenb-0.8.p2.spkg

[sagetrac-GeorgSWeber]

comment:4

I goofed it. I r-eadded "mechanize.0.1.11.zip", and in the file spkg-install, changed the version number in the corresponding line. But then, the line 6 reads "easy_install mechanize-0.1.11.tar.gz". Since there is now ...tar.gz, but "only" a ...zip, the install blows up exactly as before, if there is not internet connection.

I changed the line to end with ...zip instead, and everything went fine (on a MacIntel with OS X 10.4.11), even all doctests pass!

But since that computer has no internet connection, I wasn't able to upload that yet.

Tim, your version of the spkg also still suffers from this flaw, so I fear this ticket is back to "needs_work" again.

[TimDumol]

comment:5

I checked the spkg I posted, and I don't see any problem.

timdumol@tim-pc sagenb-0.8]$ ls src/
sagenb/                    zope.i18nmessageid-3.5.2.zip
ClientForm-0.2.10.tar.gz   zope.interface-3.6.1.tar.gz
mechanize-0.1.11.tar.gz    zope.schema-3.6.3.zip
pytz-2010h.zip             zope.testbrowser-3.8.1.tar.gz
zope.event-3.5.0-1.tar.gz

[TimDumol]

comment:6

And the contents of spkg-install:

cd src
easy_install pytz-2010h.zip
easy_install zope.i18nmessageid-3.5.2.zip
easy_install zope.event-3.5.0-1.tar.gz
easy_install ClientForm-0.2.10.tar.gz
easy_install mechanize-0.1.11.tar.gz
easy_install zope.interface-3.6.1.tar.gz
easy_install zope.schema-3.6.3.zip
easy_install zope.testbrowser-3.8.1.tar.gz
cd sagenb
python setup.py install

[sagetrac-mvngu]

comment:7

A trivial point. When uncompressed, Tim's updated spkg has this name:

[mvngu@sage sagenb]$ ls
sagenb-0.7.2.spkg  sagenb-0.7.5.1.spkg  sagenb-0.7.5.2.spkg  sagenb-0.8.p2.spkg
[mvngu@sage sagenb]$ tar -jxf sagenb-0.8.p2.spkg
[mvngu@sage sagenb]$ ls
sagenb-0.7.2.spkg    sagenb-0.7.5.2.spkg  sagenb-0.8.p2.spkg
sagenb-0.7.5.1.spkg  sagenb-0.8

Notice the directory name sagenb-0.8. It should be sagenb-0.8.p2. An updated package with directory name conforming to the Developer's Guide is available at

http://sage.math.washington.edu/home/mvngu/spkg/standard/sagenb/sagenb-0.8.p2.spkg

This is essentially Tim's spkg, but I changed the name as per above instructions, prior to tar gzip2 compressing the updated spkg.

[sagetrac-mvngu]

Reviewer: Minh Van Nguyen

[sagetrac-GeorgSWeber]

comment:8

Tim, Minh, you both are right, of course!

[sagetrac-GeorgSWeber]

Changed reviewer from Minh Van Nguyen to Minh Van Nguyen, Georg S. Weber

[sagetrac-GeorgSWeber]

comment:9

OK, using the (slightly altered) version that Minh posted from Tim's p2 spkg, I found out the following:

• this spkg solves the original problem, i.e. does not need an internet connection to build (I tested this "on top" of a full sage-4.4.1 build, forcing sagenb to be rebuilt)

• all sagenb doctests still pass

• the notebook starts up and passes some smoke tests

• the mercurial repository is clean, and the spkg is based upon the previous version 0.8.p1 (i.e. the last patch applied by Mike Hansen for some t2 issue IIRC should be still in)

• the changes made look sane (esp. about "== 0.1.11" w.r.t. the mechanize version)

• the one reviewer comment about testing has been incorporated into the SPKG.txt

I now understand the sagenb spkg better, i.e. that the file "spkg-install" is being generated, and that the dependencies under src/ are being downloaded during a standard generation of some new version of this spkg. So the discrepancy that crept in my goofed attempt just cannot occur this way. What I disliked, is that this did lead also to the following updates of certain dependencies in between sagenb-0.8.p1.spkg and sagenb-0.8.p2.spkg:

• zope.event-3.4.1.tar.gz --> zope.event-3.5.0-1.tar.gz

• zope.i18nmessageid-3.5.1.tar.gz --> zope.i18nmessageid-3.5.2.zip

• zope.interface-3.5.3.tar.gz --> zope.interface-3.6.1.tar.gz

• zope.schema-3.6.1.tar.gz --> zope.schema-3.6.3.zip

Personally, in the light of my other findings and tests above, I do not consider this as critical, assuming that sagenb does not use any "bleeding edge" functionality of the above four zope packages, that would be affected by these updates. If somebody thinks I'm wrong here, then please feel free to switch the status back to "needs_work".

[sagetrac-GeorgSWeber]

Author: Tim Dumol

[TimDumol]

comment:10

Zope very rarely makes backward incompatible changes to its API (afaik), so I think it's reasonably safe to update automatically. Any problems will be spotted during testing, anyways. Otherwise, we can change the dependencies to '==' instead of '>='.

[sagetrac-mvngu]

Merged: sage-4.4.2.alpha0