Sage 6.10.rc2 binaries test

[Volker Braun]

Now with the absolute symlinks rewritten to be relative. As before, you can find the tarballs at http://files.sagemath.org/binaries/

[Eric Gourgoulhon]

I've tested sage-6.10.rc2-Ubuntu_14.04-x86_64.tar.bz2.
Looks good to me (the Jupyter issues are fixed). Thanks.

[Henri Girard]

Hi,
I have these annoying message with ./sage -n=ipython
I does work well on ./sage -n
Any help ?
Regards
Henri


./sage -n=ipython
┌────────────────────────────────────────────────────────────────────┐
│ SageMath Version 6.10.rc2, Release Date: 2015-12-16 │
│ Type "notebook()" for the browser-based notebook interface. │
│ Type "help()" for help. │
└────────────────────────────────────────────────────────────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Warning: this is a prerelease version, and it may be unstable. ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛
Please wait while the Sage Notebook server starts...
[I 00:52:18.922 NotebookApp] Using MathJax: nbextensions/mathjax/MathJax.js
[I 00:52:18.937 NotebookApp] Writing notebook server cookie secret to
/run/user/1000/jupyter/notebook_cookie_secret
[W 00:52:19.038 NotebookApp] Terminals not available (error was No
module named terminado)
[I 00:52:19.038 NotebookApp] Serving notebooks from local directory:
/home/riri/sage
[I 00:52:19.039 NotebookApp] 0 active kernels
[I 00:52:19.039 NotebookApp] The IPython Notebook is running at:
http://localhost:8888/
[I 00:52:19.039 NotebookApp] Use Control-C to stop this server and shut
down all kernels (twice to skip confirmation).
/home/riri/sage/local/lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not
found (required by /usr/lib/x86_64-linux-gnu/libproxy.so.1)
Failed to load module:
/usr/lib/x86_64-linux-gnu/gio/modules/libgiolibproxy.so
/usr/lib/firefox/firefox: /home/riri/sage/local/lib64/libstdc++.so.6:
version `GLIBCXX_3.4.21' not found (required by /usr/lib/firefox/firefox)

[Henri Girard]

Sorry I just read your message... http... works

[Eric Gourgoulhon]

Hi,

I've tested sage-6.10.rc2-Ubuntu_15.10-x86_64.tar on a Ubuntu 15.10 computer and got the following issue:
./sage -n jupyter
results in the error message:

Please wait while the Sage Notebook server starts...

[I 11:01:15.391 NotebookApp] Using MathJax: nbextensions/mathjax/MathJax.js
[W 11:01:15.420 NotebookApp] Terminals not available (error was No module named terminado)
[I 11:01:15.421 NotebookApp] Serving notebooks from local directory: /home/eric/sage/sage-bin/SageMath
[I 11:01:15.421 NotebookApp] 0 active kernels
[I 11:01:15.421 NotebookApp] The IPython Notebook is running at: http://localhost:8888/
[I 11:01:15.421 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
/home/eric/sage/sage-bin/SageMath/local/lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not found (required by /usr/lib/x86_64-linux-gnu/libproxy.so.1)

Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libgiolibproxy.so

/usr/lib/firefox/firefox: /home/eric/sage/sage-bin/SageMath/local/lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not found (required by /usr/lib/firefox/firefox)

As a consequence, the notebook is not automatically opened in Firefox.
However, if I manually open http://localhost:8888/ everything is fine: Jupyter notebook works (including MathJax display and 3D graphics).

For sage notebook, there is no problem: ./sage -n triggers the opening of a new Firefox tab.

On the same computer, there is no such problem with a compiled version of sage 6.10.rc1: ./sage -n jupyter does open a new Firefox tab (and there is no message "version `GLIBCXX_3.4.21' not found").

Eric.

[Francois Bissey]

Fun! Your distro is built with a gcc that is more recent than the one your sage
binary has been built with - and the libstdc++ from your distro is masked by the one from
sage.

François

> --
> You received this message because you are subscribed to the Google Groups "sage-release" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sage-release...@googlegroups.com.
> To post to this group, send email to sage-r...@googlegroups.com.
> Visit this group at https://groups.google.com/group/sage-release.
> For more options, visit https://groups.google.com/d/optout.

[Henri Girard]

Yes I wrote the same problem on 16.04. I run sage -n=ipython  and then I opened firefox. I noticed for me it's even better I just created an icon with sage -n=ipython (as application) so I don't see any terminal running and I run firefox. I noticed that file can be relocated in any dir, which is a good improvment.

[Henri Girard]

yes... But not fun :)

[Henri Girard]

I noticed too, that sagemanifolds compiled perfect on the rc2 version. I
think I have gcc 5xx.

Le 18/12/2015 11:16, Francois Bissey a écrit :

[Volker Braun]

You can also run

sage --notebook=ipython --no-browser

to disable the browser auto-start

[Eric Gourgoulhon]

Le vendredi 18 décembre 2015 11:16:05 UTC+1, François a écrit :

Fun! Your distro is built with a gcc that is more recent than the one your sage
binary has been built with - and the libstdc++ from your distro is masked by the one from
sage.

What is strange is that the sage binary is called "Ubuntu 15.10" and my distro is exactly that one, with the standard gcc shipped with it, i.e. gcc 5.2.1.
 

[leif]

Eric Gourgoulhon wrote:
> Le vendredi 18 décembre 2015 11:16:05 UTC+1, François a écrit :
>
> Fun! Your distro is built with a gcc that is more recent than the
> one your sage
> binary has been built with - and the libstdc++ from your distro is
> masked by the one from
> sage.
>
>
> What is strange is that the sage binary is called "Ubuntu 15.10" and my
> distro is exactly that one, with the standard gcc shipped with it, i.e.
> gcc 5.2.1.

Not strange (although probably a flaw), but all binaries are built with
SAGE_INSTALL_GCC=yes AFAIK, so Sage is mainly built with GCC 4.9.2 which
ships with an older libstdc++.

The issue will presumably vanish when we get rid of setting
LD_LIBRARY_PATH (and using sage-native-execute here and there); until
then, you can simply create symlinks to your distro's libstdc++,
libgfortran, etc.


-leif

[Eric Gourgoulhon]

Hi,

Le vendredi 18 décembre 2015 16:46:19 UTC+1, leif a écrit :

Not strange (although probably a flaw), but all binaries are built with
SAGE_INSTALL_GCC=yes AFAIK, so Sage is mainly built with GCC 4.9.2 which
ships with an older libstdc++.

The issue will presumably vanish when we get rid of setting
LD_LIBRARY_PATH (and using sage-native-execute here and there); until
then, you can simply create symlinks to your distro's libstdc++,
libgfortran, etc.

Thanks for these explanations.
But this means that all Ubuntu 15.10 users will be facing the same issue.
Why are Sage binaries built with SAGE_INSTALL_GCC=yes instead of
using gcc of the targeted system ? With Ubuntu 15.10, there is no problem
with the latter. Note that 15.10 is the latest Ubuntu version, so it is
somehow annoying if SageMath is shipping bad binaries for it, especially
given the present discussion of making jupyter the default notebook.

Best wishes,

Eric.

[Volker Braun]

The binaries are all built with SAGE_INSTALL_GCC=yes to be as self-contained as possible. AFAIK a default Ubuntu install doesn't include a fortran compiler, for example.

The LD_LIBRARY_PATH problem is fixed in Trac #19641

[Henri Girard]

I am using daily build 16.04 and now compiling is working perfect with jupyter, there might be a time of latency between both version 15.10/16.04 and everything should be working smoothly again ?

--

[Eric Gourgoulhon]

Le mardi 22 décembre 2015 09:58:41 UTC+1, Volker Braun a écrit :

The binaries are all built with SAGE_INSTALL_GCC=yes to be as self-contained as possible. AFAIK a default Ubuntu install doesn't include a fortran compiler, for example.

The LD_LIBRARY_PATH problem is fixed in Trac #19641

OK I see.
But then there should be some warning somewhere telling the user to manually open http://localhost:8888/ and to not pay attention to the error message. Otherwise, I imagine many users will be disapointed (all the more that it worked well with Sage 6.9).

Best wishes,

Eric.