[Python-bugs-list] [Bug #113773] Python 2.0b1 selectmodule compile error under Redhat 5.2

[noreply@sourceforge.net]

Bug #113773, was updated on 2000-Sep-06 18:26
Here is a current snapshot of the bug.

Project: Python
Category: Build
Status: Closed
Resolution: Fixed
Bug Group: Platform-specific
Priority: 1
Summary: Python 2.0b1 selectmodule compile error under Redhat 5.2

Details: Environment: Redhat 5.2 running the gcc compiler that comes with the
distribution (gcc 2.7.2.3).

'make' failed when compiling selectmodule.c ('./configure' ran ok)because
the following constants were undefined:

POLLRDNORM, POLLRDBAND, POLLWRBAND

The /usr/include/sys/poll.h file does not contain these constants.
Under Redhat 6.2 the constants are defined in the /usr/include/bits/poll.h
file (which is not present in the Redhat 5.2 distribution).

I used the Redhat 6.2 poll.h file and the selectmodule compiled (under
Redhat 5.2) but got the following error when running 
'make test' (it looks like it might be related to the compile problem):

test test_poll crashed -- select.error: (9, 'Bad file descriptor')

1 test failed: test_poll



Follow-Ups:

Date: 2000-Sep-07 15:06
By: jhylton

Comment:
Please do triage on this bug.
-------------------------------------------------------

Date: 2000-Sep-14 23:08
By: tim_one

Comment:
Boosted the priority and assigned to Fred.  Don't know what you can do about an older Redhat build problem, and it's not clear that the test failure isn't simply due to mixing files across OS releases.  I'm inclined to blame it on the OS.
-------------------------------------------------------

Date: 2000-Sep-15 13:11
By: fdrake

Comment:
I've asked SourceForge if there are ways to use older versions of operating systems on their compile farm.  Until then, we'll need a volunteer with a RH5.2 system to help with this.

Stuart:  Are you willing to test a patch if I mail it to you?  Thanks.
-------------------------------------------------------

Date: 2000-Sep-15 19:50
By: fdrake

Comment:
SourceForge will try and provide a RH5.2 installation for us; now that's service!
-------------------------------------------------------

Date: 2000-Sep-20 22:33
By: fdrake

Comment:
Reducing priority since the platform is old and we still don't have access to it.
-------------------------------------------------------

Date: 2000-Sep-22 07:32
By: fdrake

Comment:
For the record, SF stills plans to attempt to provide a RH5.2 installation / "virtual environment" we can use, but the person who handles it says it will be this weekend or next week before he can get to it.
-------------------------------------------------------

Date: 2000-Oct-06 09:01
By: twouters

Comment:
I have access to a few RedHat 5.2 machines. I'll see if I can reproduce the problem there. 

(I consider the undefined values the problem, *not* the test crash. RedHat 6 uses glibc 2.1, RedHat 5 uses glibc 2.0, and those header files should not be considered portable. At best it defined the constants to values that are meaningless to the underlying library, at worst it made select.poll() delete random files from the filesystem ;-)
 
-------------------------------------------------------

Date: 2000-Oct-06 09:09
By: twouters

Comment:
Actually, this was already fixed by akuchling, in revision 2.48 of selectmodule.c (based on a patch by Neil Schemenauer.) I've confirmed that this indeed fixes the problem on RedHat 5.2: selectmodule compiles fine, and the relevant tests all pass.

-------------------------------------------------------

For detailed info, follow this link:
http://sourceforge.net/bugs/?func=detailbug&bug_id=113773&group_id=5470