>>> On 3/12/2008 at 4:31 PM, in message <[EMAIL PROTECTED]>, Carlo
Marcelo Arenas Belon <[EMAIL PROTECTED]> wrote:
> On Wed, Mar 12, 2008 at 02:20:45PM -0600, Brad Nicholes wrote:
>>   I think I am going to see some more clarification on this.  
>> --enable-static 
> doesn't seem to fit exactly with what we want to do.  According to the 
> automake docs (http://sourceware.org/autobook/autobook/autobook_85.html), 
> --enable-static and --enable-shared deals with building static or dynamic 
> libraries rather than building a statically linked or dynamically linked 
> application.  In our case what we are doing is building a statically linked 
> application.
> 
> if libganglia is a static library, it will be linked statically with
> gmond/gmetad/gmetric/gstat generating a statically linked application when
> libtool builds the target (using -static).
> 
> if all the other dependant libraries are also build statically, they will be
> linked statically as well once they are located --with-${LIBRARY}
> 

This is where my confusion is.  --enable-static does not mean build the 
application statically.  What it means is produce a statically linkable 
library.  In other words, I can define './configure --enable-static 
--enable-dynamic'.  These two options are not mutually exclusive.  What they 
mean is produce both a statically linkable and a dynamically linkable 
libraries.  That is OK if we are talking about building libganglia only.  
However what we are talking about is building applications (gmond, gmetad, 
etc.).  --enable-static as well as --enable-dynamic has no affect on how the 
gmond, gmetad, etc applications are linked to any other libraries.  In looking 
around at other projects I have seen them use --enable-static-link or 
--enable-static-exec which have the same meaning as our --enable-static-build.  

>> The only thing that --enable-static would affect is whether to build 
>> libganglia 
> as a dynamic library or as a static library.  It would have nothing to do 
> with the way gmond, gmetad, gstat or gexec are built.  I think the question 
> here is do we want to continue to support static build (ie. 
> --enable-static-build) that produces standalone binaries?  My vote would  
> be no, however if we decide to continue I think my original proposal is what 
> is needed.
> 
> Removing support for static builds wasn't the question, but as I said in my
> original post by removing the libraries in srclib we are making supporting 
> it
> more difficult.
> 
> Removing support to static binaries (which IMHO makes a lot of sense in HPC)
> just because we can't deal with the complexities we impose in ourselves
> shouldn't be an option.
> 
> Carlo
> 
> PS. I was waiting for a definition on confuse to proceed with removing it in 
> a
> clean way (as done with expat) can we have a closure on that?
> 

I think what I am saying is that we should dump libconfuse out of our SVN 
repository as well.  I have everything working already as far as being able to 
build both statically and dynamically using the --enable-static-build configure 
option.  I am all for dumping the static builds completely and therefore 
removing --enable-static-build.  But I don't think that we can rework the 
static build around the --enable-static configure option.  If you have a better 
or cleaner idea, I'm all for it.  If you are also in favor of completely 
removing static builds, just let me know and I will have it ripped out in about 
2 minutes.  :)

Brad


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Ganglia-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-developers

Reply via email to