[ADMB Users] ADMB with gcc, mingw, and/or msys - Please fix the bug identified by Dave

Sat Nov 16 10:03:55 PST 2013

On 13-11-15 11:49 PM, Jon Schnute wrote:

As far as I can tell we are using the same code from source forge and 
installing the same way.

I moved to a non re example.  To simplify I also set the environment 
string ADMB_HOME

ADMB_HOME=H:\oldave\admb\build\dist

However

H:\oldave\admb\examples\admb\catage>admb.bat catage

produces

Error: tpls.tpl not found

I can do the job with

tpl2cpp catage

followed by

H:\oldave\admb\examples\admb\catage>adcomp.bat catage

which produces

g++ -c  -O3 -D__GNUDOS__  -Dlinux -DOPT_LIB -DUSE_LAPLACE -fpermissive 
-I. -I"H:\oldave\admb\build\dist\include" 
-I"H:\oldave\admb\build\dist\contrib\include" -o catage.obj catage.cpp

and creates catage.exe which runs fine.

I suspect you are finding the wrong library somewhere.  You could edit 
the adcomp.bat file to add a -v

g++ -c  -v -O3 -D__GNUDOS__  -Dlinux -DOPT_LIB -DUSE_LAPLACE 
-fpermissive -I. -I"H:\oldave\admb\build\dist\include" 
-I"H:\oldave\admb\build\dist\contrib\include" -o catage.obj catage.cpp

and see where it finds all the libraries.

> Very interesting! You're obviously a better picker of Gnu software than I
> am. Can you please give me the precise link you used to download the Gnu
> software? Also, were you working with source code from ADMB 11.1 or the more
> current version from the SVN repository?
>
> Incidentally, I used the Gnu installer 'mingw-get-setup.exe' available at:
>
> http://sourceforge.net/projects/mingw/files/Installer/
>
> I then used this to install the three packages mingw32-base,
> mingw32-gcc-g++, and msys-base. The fact that you and I get different
> results shows that it matters which software is used. (By the way, I used an
> example without random effects, but surely that shouldn't be relevant.)
>
> I have another question for the ADMB community. The ADMB Downloads site
>
> http://code.google.com/p/admb-project/downloads/list
>
> has the file 'admb-11.1-windows-mingw-32bit.zip'. Can anyone tell me
> precisely which Gnu software was used to build this binary image? Someone
> must know because someone must have built it.
>
> Thanks. I think we're getting close.
>
> Jon
>
> -----Original Message-----
> From: dave fournier [mailto:davef at otter-rsch.com]
> Sent: November-15-13 2:58 PM
> To: Jon Schnute; users at admb-project.org
> Subject: Re: [ADMB Users] ADMB with gcc, mingw, and/or msys - Please fix the
> bug identified by Dave
>
> On 13-11-15 01:54 PM, Jon Schnute wrote:
>
> No it compiled fine. It was the tpl2rem step that failed.
> Since I was interested in the link step with g++ I used a different version
> if ADMB to produce the polio.cpp and polio.htp files.
> The compile and link using these files ran fine.  So I guess I have no idea
> why you got the winmain problem.
>
>
>
>> Thanks Dave! That's very helpful and it replicates my own experience,
>> although you don't say precisely how the attempt to build the 'polio'
>> example failed. (I wonder if it involved a mysterious reference to
>> 'WinMain at 16' at the link step.)
>>
>> Please note that my request has nothing to do with R. All I want is a well
>> defined procedure for building ADMB on Windows with msys/gcc. Your example
>> and my own experience confirm that it is possible to build the ADMB binary
>> library successfully by that route. And I'm really pleased to learn that
>> this library actually is OK. However, you've demonstrated that the current
>> version of ADMB has a bug in its batch scripts in the \bin directory. As
> you
>> say, the command
>>
>> admb -r polio
>>
>> failed, yet you used manual scripts to make it work.
>>
>> Can someone please fix this bug? I'm not sure how this would work, but
>> perhaps you need to send the detailed scripts that actually worked for you
>> to the right person (Johnoel?).
>>
>> Incidentally, I suggested Rtools only as a possibly reliable source for
>> mingw/gcc, given that Arni has used it in the past. I realize that many
> ADMB
>> users have probably abandoned the Windows OS, but you still have a
> sizeable
>> user community that uses that platform. All I'm asking is that Windows
> users
>> should be able to download ADMB source, build it, and use it like everyone
>> else. Unfortunately, because Windows is a crippled OS, we need to take
>> seriously the source for Unix tools on that platform. From your
> experience,
>> it sounds as if that shouldn't be a big deal, although the batch scripts
>> need to be fixed.
>>
>> I'm trying to be helpful. Once I have a procedure that actually works,
> I'll
>> be glad to document it if people are interested. Please note that we need
> to
>> consider both 32- and 64- bit versions of the software. I've noticed that
>> several msys sources are available on the Internet, so the documentation
>> would need to be explicit about that.
>>
>> Best wishes,
>>
>> Jon
>>
>> -----Original Message-----
>> From: users-bounces at admb-project.org
> [mailto:users-bounces at admb-project.org]
>> On Behalf Of dave fournier
>> Sent: November-14-13 3:53 PM
>> To: users at admb-project.org
>> Subject: Re: [ADMB Users] ADMB with gcc, mingw, and/or msys
>>
>>
>> I was curious to see how hard this is.
>>
>> I downloaded the current mingw compiler and msys tools from source forge.
>> I modifed my path to include the three directores for gcc msys tools and
>> admb as below.
>>
>> PATH=H:\oldave\admb\build\dist\bin;c:\MinGW\msys\1.0\bin\;c:\MinGW\bin;
>>
>> I changed to the admb directory and typed make.  ADMB was built
>> successfully.
>>
>> then I tried to compile the polio example
>>
>> The command
>>
>>                admb -r polio
>>
>> did not work for me.  Since I was really interested in whether the library
>> is correct and don't like those overly complicated scripts anyway I built
>> the cpp and htp by hand and then compiled
>> it with   mygccs.bat where mygccs.bat contained
>>
>> g++.exe -Dlinux -DUSE_LAPLACE polio.cpp -o polio.exe -I.
>> -Ih:/oldave/admb/build/dist/include -Lh:/oldave/admb/build/dist/lib -ladmb
>>
>>
>> This built the polio.exe which ran successfully.
>>
>> I can't imagine  the average R user getting anywhere with this.
>>
>>
>>
>>
>>
>> _______________________________________________
>> Users mailing list
>> Users at admb-project.org
>> http://lists.admb-project.org/mailman/listinfo/users
>>
>>
>>
>
>
>