--with-blas/lapack being ignored

[Thomas Treichl]

Ben Abbott schrieb:
> On Wednesday, December 05, 2007, at 10:26AM, "Thomas Treichl" <Thomas.Treichl at gmx.net> wrote:
>> Ben Abbott schrieb:
>> <SNIP>
>>> Running fntests still crashes.
>>>
>>> [...]
>>>   src/octave/scripts/time/datestr.m ...................... PASS   32/32  
>>>   src/octave/scripts/time/datevec.m ......................panic: Bus 
>>> error -- stopping myself...
>>> Bus error
>>>
>>> Which may be duplicated with the simpler ....
>>>
>>> strptime ('09/13', '%m/%d')
>>> panic: Bus error -- stopping myself...
>>> attempting to save variables to `octave-core'...
>>> save to `octave-core' complete
>>> Bus error
>>>
>>> However,  ....
>>>
>>> strptime(date,'%d-%b-%Y')
>>> ans =
>>> {
>>>   usec = 0
>>>   sec = 0
>>>   min = 0
>>>   hour = 0
>>>   mday =  5
>>>   mon =  11
>>>   year =  107
>>>   wday =  3
>>>   yday =  338
>>>   isdst = 0
>>>   zone = EST
>>> }
>>>
>>> Might the problem lie in one of octave's dependencies? 
>>>
>>> Ben
>> Only the datevec dependent problem: I expect YES Ben, this still seems to me 
>> being a compiler or a dependency problem. I was asking you to try the simple 
>> command 'strptime('09/13','%m/%d')' because I somehow sensed that it could be 
>> this problem - a bug has been solved for the Apple strptime function 6 weeks ago 
>> and it somehow seems to me now that you either are not using XCode's original C 
>> library *or* that another dependency that uses Apple's C library isn't compiled 
>> bug-free -> 'panic: Bus error'. Please correct me everybody if this can't be true...
>>
>> So either there is a compiler problem or a dependency problem (both must be on 
>> the Fink side of life as it seems to me) because it does work for Marius's HMUG 
>> binary version of Octave and for Octave.app too. However, at this point I don't 
>> know how I could help here to find out what is going wrong, I just can tell you 
>> about my thoughts and about my experiences...
>>
>>   Good luck,
>>   Thomas
>>
> 
> Thomas, what exactly was fixed 6 weeks ago? ... octave, an XCode library, or something else?

The strptime C function of Apple's C library does not return that value(s) that 
we expect it should return. That's why a workaround has been implemented for 
this and instead of using the strptime function from Apple's C library Mac users 
should use the ./liboctave/strptime.c function instead (or at least on every 
system this problem appears). This will be found out automatically (see below).

> In either case, please let me know how I'd verify I have the fix.

There is a configure test if you run './configure' for Octave then watch the 
output or do this

   bash$ ./configure <all your options> | grep "strptime"

and you should get something like

   checking for strptime... yes
   checking whether strptime is broken... yes

what do you see? Are you familiar with debugging?

> I'm working wth the 2.9.17 release of octave, but can easily upgrade. 

Not necessary, try to find out that specific problem with the 2.9.17 sources. 
Another upgrade might displace the current bug-problem maybe to another problem 
and we can somehow see this clearer now. Even if 2.9.18 comes out in a few 
moments, continue finding the problem with the 2.9.17 sources until found.

> My XCode is 3.0.

Wow, you're fast ;) But I think this is ok because it seems to me still being 
the same GCC compiler version as in XCode's 2.4.0 (mine).

> gcc --version 
> i686-apple-darwin9-gcc-4.0.1 (GCC) 4.0.1 (Apple Inc. build 5465)
> Copyright (C) 2005 Free Software Foundation, Inc.
> This is free software; see the source for copying conditions.  There is NO
> warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
> 
> It is my understanding that some/many of Fink's packages are compiled using 4.2.2 instead of 4.0.1. This is due to a desire to switch form g95 to gfortran. Is it possible that this problem follows from mixing compiler versions?

Maybe, I don't know - it might be or it might be not, that's what I meant before 
that I can't check for this problem and can't help here.

   Thomas