Re: [Libmesh-devel] current build system vs. automake

[Kirk, B. (JSC-EG311) <ben...@na...>]

On 2/22/12 9:40 AM, "Paul T. Bauman" <ptb...@gm...> wrote:

>> Thanks...  At this point I've got an alpha implementation that builds itself
>> and all the contributed programs, and links one example.  I want to spend
>> some
>> time working on how we treat the examples though so...
>>

https://libmesh.svn.sourceforge.net/svnroot/libmesh/branches/libmesh.automake


>> I'm still working feverishly on it so I wouldn't make any major changes ATM,
>> but I would appreciate if you guys wanted to take a look and comment or
>> complain.
>
> One thing I would like to tackle is add some versioning info so that we can
> update our libmesh.m4 file to use it so we can check minimum version
> requirements (in addition to cleaning up the libmesh.m4 file to use this
> infrastructure; then that could be shipped with the library).

all the adaptivity and introduction examples run properly from both 'make check' and 'make distcheck'

For some reason you have to do 'make' before 'make check' - there is a dependency which isn’t propagating all the way.  I'll get that later.

In the mean time I am going to look at the pkg-config stuff.

Note that when you do 'make install' you get all the examples in a nice, self explanatory directory tree.  My intent is to have each of those directories stand alone, so you'll see while in the source we may have share source between examples those files will get installed as duplicates.  This is intentional, so someone can just grab a directory, go in there, and start modifying it on their own without knowing intro_ex4 needs a source from intro_ex3.  So if you see some weirdness in examples/Makefile.am that is probably why.

At this point  I am going to stop messing with the examples.  I think the general framework is good but it could use some polishing - perhaps a shared set of utilities to replace what is now duplicate code?  Paul, if you are interested in helping out either porting more examples or maybe removing some of the duplication in the current run.sh's would be awesome.  Just remember to do a 'make distcheck ' to catch anything tricky.

I am going to move to getting 'libmesh-config' to do the right thing on an installed source tree and then it will just be commenting, fixing make doc, and all the other little make targets we've added here and there.

-Ben