[OS] Another data point for OS / ASL under Msys / cl

Kipp Martin kipp.martin at chicagogsb.edu
Sat Sep 15 13:19:20 EDT 2007


Hi Bob:

> If ASL support is dropped, what do we lose?  Is there a way ASL can still be
> made available, but without some of the desirable features of COIN-OR's
> project setup?

To answer the first question, I think the OS project loses a lot without 
ASL. Without ASL we can't do any processing of nl files or use 
amplClient.  I think the answer to the  second question above is no. 
Right now you really do build ASL "separately" from the COIN projects 
and that is causing the problem. For example, the COIN-OR BuildTools 
configure script is used to build the COIN-OR project Makefiles but ASL 
is built with the Makefiles that it downloads with -- there is no 
coordination. So I see the problem as ASL not being part of the COIN-OR 
project setup.  If it were, then when we use the debug option for the 
COIN-OR configure script the debug option would get put into the ASL 
Makefile. That does not happen now.

> 
>> Sorry, I have no idea.  NlpProblemDef is not a symbol 
>> in our code -- it must be in OSI.
> 
> Does this make any sense?

Not really. When I build the OS project using --enable-debug with Knitro 
I get the crash described in the previous email I sent to you. When I 
build the OS project again using --enable-debug without Knitro 
everything compiles, links, and passes the unitTest. So it would seem 
Knitro causes the problem since there is no problem when the project 
builds without Knitro. I will investigate further.

Regards,


> 
> -- Bob
> 
> 
>> -----Original Message-----
>> From: Kipp Martin [mailto:Kipp.Martin at ChicagoGSB.edu]
>> Sent: Saturday, September 15, 2007 1:10 AM
>> To: optimizationservices at googlegroups.com
>> Cc: Jun Ma @ NWU; Lou Hafer; Robert Fourer
>> Subject: Re: [OS] Another data point for OS / ASL under Msys / cl
>>
>> Hi Lou:
>>
>> I am able to replicate your error when running
>>
>> ./configure --enable-debug.
>>
>> This is actually a generic problem and not just an OS problem.  It seems
>> like one COIN-OR  needs to resolve as this will affect (I think) any
>> project that tries to link to ASL.
>>
>> 1) I checked out the Cbc project independently of OS and got the same
>> error.
>>
>> 2) It is an ASL problem. If you build without ASL it works.
>>
>> Now here is what I do not understand. I did a
>>
>> make distclean
>>
>> and went back and modified the ASL makefile.vc to be
>>
>> CFLAGS = -nologo -Ot1 -MTd
>>
>> Then I did a completely new configure. However, after doing this I still
>> get the same error, namely
>>
>> LIBCMT.lib(calloc.obj) : error LNK2005: _calloc already defined in
>> LIBCMTD.lib(dbgheap.obj)
>>
>> Do you have any idea how LIBCMT.lib is getting linked in there after
>> changing the ASL makefile as described above?
>>
>>
>> At this point I am tempted to simply drop ASL support from OS, I have
>> spent such a ridiculous  amount of time on it over the last year or so.
>>
>> Cheers,
>>
>>
>>
>>
>>> 	An optimised build went through without errors.  `make test' was
>>> superficially ok (i.e., no core dump).  I didn't go over the output in
>> detail,
>>> but OS claimed it passed its unit test.  Kipp, you mentioned that you
>> just type
>>> `./configure' to do the build --- this gives an optimised build.
>>>
>>> 	The debug build still runs into trouble trying to link cbc.exe.  I
>> think
>>> if you use the command `./configure --enable-debug' you will be able to
>>> replicate the problem.  (My actual configure command is considerably
>> longer and
>>> built by a script, but this should do it.)  Coin will specify -MTd as
>> the
>>> multi-thread library, while ASL will continue to use -MT.  The two
>> collide at
>>> the attempt to link cbc.exe.
>>>
>>> 	I ran the debug build to the cbc error, then switched to the
>> optimised
>>> build, and didn't have time to patch the debug build and check for the
>> __matherr
>>> problem. I'll try that next time I'm working at home.
>>>
>>> 							Lou
>>>
>>> _______________________________________________
>>> OS mailing list
>>> OS at list.coin-or.org
>>> http://list.coin-or.org/mailman/listinfo/os
>>>
>>> >>>
>>
>> --
>> Kipp Martin
>> Professor of Operations Research
>> and Computing Technology
>> Graduate School of Business
>> University of Chicago
>> 5807 South Woodlawn Avenue
>> Chicago, IL 60637
>> 773-702-7456
>> kipp.martin at chicagogsb.edu
>> http://gsbkip.chicagogsb.edu
>> http://www.coin-or.org
> 
> 


-- 
Kipp Martin
Professor of Operations Research
and Computing Technology
Graduate School of Business
University of Chicago
5807 South Woodlawn Avenue
Chicago, IL 60637
773-702-7456
kipp.martin at chicagogsb.edu
http://gsbkip.chicagogsb.edu
http://www.coin-or.org

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Optimization Services (OS)" group.
To post to this group, send email to optimizationservices at googlegroups.com
To unsubscribe from this group, send email to optimizationservices-unsubscribe at googlegroups.com
For more options, visit this group at http://groups.google.com/group/optimizationservices?hl=en
-~----------~----~----~----~------~----~------~--~---



More information about the OS mailing list