[reportlab-users] Install problem with Reportlab 2.0

Andy Robinson andy at reportlab.com
Fri Oct 6 14:04:42 EDT 2006

You are right. FWIW our official position is probably still that "we do not support distutils" - our documentation tells people to unzip the package and add to sys.path, and rl-accel is in theory "optional". 

The setup.py was a result of several contributions by linux users over years and has no real "owner" here, and we have either tended to use prebuilt DLLs on Windows or to compile by hand on many servers.  We've ben trusting others that certain things are needed, instead of testing. 

 I think we probably can improve this on the next release some time this quarter - with the arrival of VMWare it's a lot easier for us to test the installation on several standard Linux platforms, and we would like to offer setuptools and egg support.    But we'd have to add some other much bigger extensions too to make it simple for everyone. 

- Andy

-----Original Message-----
From: Parsifal Herzog <parsifal.herzog at gmail.com>
Date: Fri, 06 Oct 2006 12:21:38 
To:reportlab-users at reportlab.com
Subject: Re: [reportlab-users] Install problem with Reportlab 2.0

On Friday 06 October 2006 05:23, Robin Becker wrote:
> here is a patched version of setup.py; can you try this with and
> without the additional arguments
> --rl_accel=0
> --rl_accel=1
> etc?
> There was a complete borkeness in the version checking had no .c on
> the filename :(

The new setup.py failed with:

[parz at lobsang reportlab]$ python setup.py build
Traceback (most recent call last):
  File "setup.py", line 279, in ?
  File "setup.py", line 193, in run
    RL_ACCEL = _find_rl_accel()
  File "setup.py", line 89, in _find_rl_accel
  File "setup.py", line 66, in _cmp_rl_accel_dirs
    return cmp(_rl_accel_dir_info(b),__rl_accel_dir_info(a))
NameError: global name '__rl_accel_dir_info' is not defined
[parz at lobsang reportlab]$

I changed line 66 to
    return cmp(_rl_accel_dir_info(b),_rl_accel_dir_info(a))

and the install finally succeeded.

Thanks for the quick response. BTW, IMHO, the setup.py script is awfully 
complicated, and makes a lot of guesses and assumptions: such code is 
very fragile, yet it is deployed untested, probably because of the 
difficulty for developers to obtain a clean install environment. Why 
not just deliver rl_accel with Reportlab (it is very small,) and save 
all that failure-prone and unreliable testing in setup.py?

reportlab-users mailing list
reportlab-users at reportlab.com

More information about the reportlab-users mailing list