Re: [Matplotlib-users] matplotlib 0.85 on python 2.4 installation problem : "cannot find tcl/tk headers. giving up."

[Christian K. <ck...@ho...>]

Eric Emsellem wrote:
> Hi,
> trying to install matplotlib 0.85 on python 2.4 (Suse 10.0), I get :
> 
> ===> cannot find tcl/tk headers. giving up.
> 
> this is annoying and I cannot find much on the web to help me there. I
> have checked my libraries and soft (including the "devel") and think I
> have everything updated right.
> 
> It is probably a simple pb, so if anybody has a  hint there;
> 
> thanks in advance.
> 
> Eric
> P.S.: by the way, upgrading my Linux (Suse 10.0) and therefore getting
> python 2.4, I have now to reinstall everything from scratch (python
> modules) and this is a REAL pain (for example ATLAS which takes ages and
> so on). It means for me that I cannot work before having all back to
> normal (all modules working such as ppgplot, gnuplot-py, Numeric,
> numarray, Scipy, matplotlib, etc, etc). And I am not even sure this will
> compile with Python 2.4. Anybody knows a way out of this?

I was wrong about the tk/tcl issues. It is not solved by installing the devel
rpms. So you could do what Arndt proposed or live without the tk backend by
setting BUILD_TKAGG=0.
Anyway I just wanted to tell that building scipy from svn and matplotlib 0.85
was really straightforward on SuSE 10.0 with gcc4 and python2.4. Concerning the
ATLAS libs: I built them once some years ago and since that time I used them on
every freshly installed system without problems as long as the architecture was
the same (P4 32 bit). Probably the optimization suffers a bit by doing so but I
don't care that much for speed. If you like I can send you the ATLAS binaries.

Regards, Christian