[SciPy-Dev] import error in scipy.stats on RH3 32-bit
Thu Jun 3 18:28:05 CDT 2010
I've just found the problem, and it had nothing to do with
checkins that were being made to the 0.8.x branch.
After spending may hours on this problem, I'm now very
happy that I found the problem. It tured out that the
machine I use to build scipy on RH3 32-bit had it's
clock set in the past. So the new brach source tarball
I made (which is being used on all the build machines),
had timestamps which appeared to be in the future on
that machine. I'm not sure why exactly timestamps in
the future can cause problems when building scipy,
but after setting the clock on the machine things work
fine now. So it was only by accident this happend on
RH3 32-bit, it might have been any other system as well.
On Wed, Jun 2, 2010 at 10:29 PM, Ilan Schnell <firstname.lastname@example.org> wrote:
> Not yet. I'll look more into it tomorrow. :-)
> - Ilan
> On Wed, Jun 2, 2010 at 10:19 PM, Charles R Harris
> <email@example.com> wrote:
>> On Wed, Jun 2, 2010 at 9:03 PM, Ilan Schnell <firstname.lastname@example.org> wrote:
>>> Hello Chuck,
>>> yes 6446 works. Actually, as the error indicates, the unresolved
>>> symbol in is linalg/clapack.so, it just happened that during my
>>> testing the stats package was imported first, so I initially thought
>>> the error was there.
>>> However, something has changed between 6446 and 6476, as
>>> I wasn't seeing this error before. Looking at the revision log of
>>> the 0.8.x branch, but I cannot see any obvious. And I'm also
>>> puzzled why this only happens on one particular platform.
>>> To make sure the build environment hasn't changed, I rebuild 6446
>>> on the same system, and it still works.
>> I hate to ask this of anyone, but... could you determine which revision
>> caused the problem?
>> Sadistical Chuck
>> SciPy-Dev mailing list
More information about the SciPy-Dev