Opened 5 years ago

Closed 5 years ago

Last modified 5 years ago

#17347 closed defect (invalid)

matplotlib 1.3.1 breaks Sage make build

Reported by: darij Owned by:
Priority: major Milestone: sage-6.5
Component: PLEASE CHANGE Keywords: matplotlib, make, install
Cc: Merged in:
Authors: Reviewers: François Bissey, Volker Braun
Report Upstream: N/A Work issues:
Branch: Commit:
Dependencies: Stopgaps:

Description

https://www.dropbox.com/s/geklcqrouaejj2k/mpl-fail.log?dl=0

Also, sage doesn't want to build without an internet connection because of matplotlib. Not sure if this is correct behavior.

Change History (10)

comment:1 Changed 5 years ago by jdemeyer

Which version of Sage is this?

comment:2 Changed 5 years ago by darij

The newest one, 6.4.

comment:3 Changed 5 years ago by vbraun

Building shall not require internet. This is also kind-of tested on the buildbot my poisoning the proxy variables.

comment:4 Changed 5 years ago by darij

Building does require internet: it breaks with a socket exception if I am offline. Good to hear that this is not supposed to be so, but it is a bug.

comment:5 Changed 5 years ago by vbraun

I think there is something wrong with your installation, it can't find numpy but we do install numpy before matplotlib. Also, your log is multiple concatenated logs. Try "make distclean && make"

comment:6 Changed 5 years ago by fbissey

BUILDING MATPLOTLIB
            matplotlib: yes [1.3.1]
                python: yes [2.7.8 (default, Jul 19 2014, 08:57:06)  [GCC
                        4.6.3]]
              platform: yes [linux2]

REQUIRED DEPENDENCIES AND EXTENSIONS
                 numpy: yes [version 1.8.1]
              dateutil: yes [using dateutil version 2.2]
               tornado: yes [tornado was not found. It is required for the
                        WebAgg backend. pip/easy_install may attempt to
                        install it after matplotlib.]
             pyparsing: yes [using pyparsing version 2.0.1]
                 pycxx: yes [Couldn't import.  Using local copy.]
                libagg: yes [pkg-config information for 'libagg' could not
                        be found. Using local copy.]
              freetype: yes [version 17.1.11]
                   png: yes [version 1.2.51]

It says tornado was not found. numpy is found, and tornado is listed as a dep of ipython but not matplotib. Which is funny because I thought tornado was included in sage because it was unbundled from matplotlib in version 1.3 and I am thought it was part of the matplotlib ticket. If not I am considering myself negligent because I spent quite some time reviewing the ticket - including having to get pkg-config included for it.

comment:7 Changed 5 years ago by fbissey

Ha now I understand your comment Volker. It is rather incriminating that in the first install attempt only tornado was not found (and it is a proper dep of matplotlib I didn't look properly the first time) but in the last attempt numpy was not found. I think someone messed up their install by that stage.

comment:8 follow-up: Changed 5 years ago by darij

So there is a dependency of matplotlib which is not scheduled for installation before matplotlib by the makefile? Or is the tornado thing a red herring?

Anyway, thank you -- make distclean fixed it. Feel free to drop this ticket if there is nothing in here that remains to be fixed.

comment:9 in reply to: ↑ 8 Changed 5 years ago by fbissey

  • Resolution set to invalid
  • Reviewers set to François Bissey, Volker Braun
  • Status changed from new to closed

Replying to darij:

So there is a dependency of matplotlib which is not scheduled for installation before matplotlib by the makefile? Or is the tornado thing a red herring?

A red herring it was at the top of the file you put on dropbox, may be it wasn't done the first time you installed MPL on 6.4.betaX but it is sorted now.

Anyway, thank you -- make distclean fixed it. Feel free to drop this ticket if there is nothing in here that remains to be fixed.

I don't think so. I think your install was in a unhappy state for some reason. the distclean restored it a clean and happy state.

I personally think this should be closed.

comment:10 Changed 5 years ago by jdemeyer

For the record, both numpy and tornado are listed as explicit dependencies of matplotlib in build/deps. So I agree that there is no bug here.

Note: See TracTickets for help on using tickets.