Re: [Teem-users] make system simplification

[James B. <bi...@cs...>]

> Okay, so I agree that "configure" is not a must-have- its more that some 
> people just expect it to be there, and get uneasy when dealing with 
> software that doesn't conform to expectations.  The exact same can be 
> said for the people who adore cmake- for Teem its not going to be doing 
> anything fundamentally novel, but these people just get all twitchy if 
> software can't be built with it.

Using CMake would require all the SCIRun user base to install CMake.  It 
also means that you would need to compile CMake for exoctic systems 
(like AIX), that perhaps don't have a port.  It was written in C partly 
for portability, right?

> So James, it looks like in your example you're already setting a 
> per-build target directory using TEEM_DEST.  Would other people be too 
> annoyed if setting TEEM_DEST or a variable like it was mandatory?

Actually, I only set it for the debug build.  I take the default for the 
other one.  I wouldn't mind having to set it.  I would just want the 
build system to make the lower directories like include/bin/lib for me, 
so I don't have to try and fiddle with that myself.

James

> On May 21, 2005, at 1:38 PM, James Bigler wrote:
> 
>> I'm not sure what configure will buy you except more maintenance.  I 
>> kind of like the per system .mk files.  It's easy to change them, you 
>> know where all the system dependent stuff is, it's easy to override 
>> them with environment variables.  I have scripts that I use to build 
>> debug and optimized versions of teem for 3 different platforms.  It 
>> was really pretty easy and saves me from having to remember all the 
>> flags to build it just the way I want.
>>
>> James
>>
>> Here's an example:
>>
>> setenv TEEM_ZLIB
>> setenv TEEM_PNG
>> setenv TEEM_BZIP2
>>
>> # Compile a debug version
>> setenv TEEM_ARCH linux.32
>> setenv OPT_CFLAG -g
>> setenv TEEM_DEST /home/sci/bigler/pub/src/teem/linux.debug
>>
>> gmake clobber
>> gmake -j 3
>> gmake
>> set debugbuild=$?
>>
>> # Compile the default, optimized version
>> unsetenv TEEM_DEST OPT_CFLAG
>>
>> gmake clobber
>> gmake -j 3
>> gmake
>> set optbuild=$?
>>
>> # Print out the results of the compilation
>> if ($debugbuild == 0) then
>>   echo "Debug build succeeded"
>> else
>>   echo "Debug build failed"
>> endif
>>
>> if ($optbuild == 0) then
>>   echo "Opt build succeeded"
>> else
>>   echo "Opt build failed"
>> endif
>> #-----------------------------------------------
>>
>> Michael Callahan wrote:
>>
>>> I think I'll configurize it eventually.  I've just never done it before
>>> so my configure scripts might be bad.  I was hoping someone who was more
>>> familiar with that might do it first :)
>>>   Michael
>>> On Fri, 2005-05-20 at 20:08 -0400, Gordon Kindlmann wrote:
>>>
>>>> hello,
>>>>
>>>> I'm considering doing away with all the per-architecture directories 
>>>> ("darwin.32", "darwin.64", "linux.32", ...) that Teem comes with 
>>>> because its getting out of hand.  I'd rather have something like the 
>>>> TEEM_DEST variable be the norm for how the makefiles are used.  
>>>> TEEM_DEST plays the role of the "--prefix=/blah" flag to GNU's 
>>>> configure script (which someone really ought to learn enough about 
>>>> to build Teem that way). The "lib", "include", and "bin" 
>>>> sub-directories of TEEM_DEST are where things are put by the make 
>>>> system.  This is already how SCIRun uses Teem.
>>>>
>>>> This change would mean that if you want to use the same Teem 
>>>> directory tree on multiple platforms, you have an 
>>>> architecture-specific TEEM_DEST-ish thing, just like you have an 
>>>> architecture-specific TEEM_ARCH.  TEEM_ARCH would still be required 
>>>> as it is now.  I might consider having a default setting for 
>>>> TEEM_DEST, but I'd like to avoid cleverness.
>>>>
>>>> Tell me what you think.  In the next month I'd like to fix the big 
>>>> things not working perfectly in Teem: makefile stuff, 
>>>> nrrdSpatialResample (haven't heard any complaints about my proposed 
>>>> change), and non-linear least squares tensor estimation (actually 
>>>> that's not even started).
>>>>
>>>> Gordon
>>>>
>>>>
>>>>
>>>> -------------------------------------------------------
>>>> This SF.Net email is sponsored by Oracle Space Sweepstakes
>>>> Want to be the first software developer in space?
>>>> Enter now for the Oracle Space Sweepstakes!
>>>> http://ads.osdn.com/?ad_id=7412&alloc_id=16344&op=click
>>>> _______________________________________________
>>>> teem-users mailing list
>>>> tee...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/teem-users
>>>
>>> -------------------------------------------------------
>>> This SF.Net email is sponsored by Oracle Space Sweepstakes
>>> Want to be the first software developer in space?
>>> Enter now for the Oracle Space Sweepstakes!
>>> http://ads.osdn.com/?ad_id=7412&alloc_id=16344&op=click
>>> _______________________________________________
>>> teem-users mailing list
>>> tee...@li...
>>> https://lists.sourceforge.net/lists/listinfo/teem-users
>>
>>
>>
>> -------------------------------------------------------
>> This SF.Net email is sponsored by Oracle Space Sweepstakes
>> Want to be the first software developer in space?
>> Enter now for the Oracle Space Sweepstakes!
>> http://ads.osdn.com/?ad_id=7412&alloc_id=16344&op=click
>> _______________________________________________
>> teem-users mailing list
>> tee...@li...
>> https://lists.sourceforge.net/lists/listinfo/teem-users
> 
> 
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by Oracle Space Sweepstakes
> Want to be the first software developer in space?
> Enter now for the Oracle Space Sweepstakes!
> http://ads.osdn.com/?ad_id=7412&alloc_id=16344&op=click
> _______________________________________________
> teem-users mailing list
> tee...@li...
> https://lists.sourceforge.net/lists/listinfo/teem-users