This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

classification
Title: Python 2.5b2 fails to build on Solaris 10 (Sun Compiler)
Type: Stage:
Components: Build Versions: Python 2.5
process
Status: closed Resolution: fixed
Dependencies: Superseder:
Assigned To: Nosy List: akuchling, christian.heimes, nnorwitz, ostkamp
Priority: normal Keywords:

Created on 2006-07-25 20:34 by ostkamp, last changed 2022-04-11 14:56 by admin. This issue is now closed.

Files
File name Uploaded Description Edit
python_build.tar.bz2 ostkamp, 2006-07-25 20:34 Python 2.5b2 build log (tar/bzip2)
Messages (7)
msg29278 - (view) Author: Guido Ostkamp (ostkamp) Date: 2006-07-25 20:34
Hello,

I tried to build Python 2.5b2 on Solaris 10 with the
Sun Workshop Compiler, which fails.

Please see the attachment for a complete log in several
stages. Here is a short summary:

*
build/temp.solaris-2.10-sun4us-2.5/libffi/include/ffi.h
contains illegal C-syntax __attribute(...)

This might be legal for GCC, but not for other compilers. 

* Even if this is removed, there are other fatal
compiler errors in the _ctypes module later on.

* The build process seems to invoke _cursesmodule.c.
This contains the use of a function  mvwgetnstr(). This
function does not exist in /usr/lib/libcurses.so.

* The build process fails to link with libcurses.so.

The Python 2.4.3 release fails for the same reasons
regarding the curses problems, but the _ctypes stuff
did not exist in this form. 

I would appreciate, if these bugs could be fixed.

Regards,

Guido
msg29279 - (view) Author: Neal Norwitz (nnorwitz) * (Python committer) Date: 2006-07-26 04:11
Logged In: YES 
user_id=33168

Can you provide a patch to address these problems?

I don't know that any developers have access to the sun
workshop compiler.  We have a Solaris 10 box, but it only
has gcc AFAIK.
msg29280 - (view) Author: A.M. Kuchling (akuchling) * (Python committer) Date: 2006-07-26 17:28
Logged In: YES 
user_id=11375

The _cursesmodule mvwgetnstr() problem was reported as bug
#1471938 and should now be fixed on both the 2.4 and 2.5
branches.

msg29281 - (view) Author: Guido Ostkamp (ostkamp) Date: 2006-07-26 20:52
Logged In: YES 
user_id=1028306

Hello,

@nnorwitz:

Unfortunately I cannot provide a patch, but can give some hints:

If you (or the developers) have a Solaris 10 box, they can
download the Sun Studio 11 which includes the Sun Compiler
for free from the URL:
<http://developers.sun.com/prodtech/cc/downloads/index.jsp>

Then you should be able to verify the problems yourself
(though I am using an older version of the compiler from the
times when it was commercial).

Regarding the _ctypes module problems:

The module contains a libffi which nowadays appears to be
part of the GNU GCC. This lib was never made to work with
anything else but GCC, as it seems. Even older versions do
not compile.

Therefore, if this module is not really necessary, it should
be removed or at least deactivated by the configuration if
not using GCC. It also seems, that other modules like
_struct appear to be depending on this, if I'm right.

Regarding the curses problem:

I can workaround this if changing the mvwgetnstr() to
mvwgetstr() with omitting the last parameter. However this
should be configured automatically.

Futhermore there was no linking of the correct library
(/usr/lib/libcurses.so). I had to activate the curses entry
in Modules/Setup manually and to call make two more times. 

Regarding GCC on Solaris: I tried with GCC but it failed on
Solaris 10 as well. I will open a new fault report for the
GCC problems.

@akuchling:

You say the curses bug should be fixed in 2.4 and 2.5. It
isn't. I tried 2.4.3 yesterday as well and ran into the very
same problem. Or do you mean it was changed more recently?

msg29282 - (view) Author: Neal Norwitz (nnorwitz) * (Python committer) Date: 2006-07-27 03:20
Logged In: YES 
user_id=33168

Andrew Kuchling just checked in the change to fix the
problem.  The only way you would be able to verify it is if
you checkout from SVN head for 2.5 or release24-main for 2.4.

The only module that should depend on libffi is ctypes.  If
struct doesn't compile that is a different problem.
msg59789 - (view) Author: Christian Heimes (christian.heimes) * (Python committer) Date: 2008-01-12 04:58
Guido, have you tested the build on your box? I'm going to close the bug
after two week unless you report a new problem. (2008-01-26)
msg68414 - (view) Author: A.M. Kuchling (akuchling) * (Python committer) Date: 2008-06-19 17:16
Closing, per the last comment.
History
Date User Action Args
2022-04-11 14:56:19adminsetgithub: 43721
2008-06-19 17:16:25akuchlingsetstatus: pending -> closed
resolution: fixed
messages: + msg68414
2008-01-12 04:58:53christian.heimessetstatus: open -> pending
nosy: + christian.heimes
messages: + msg59789
2006-07-25 20:34:02ostkampcreate