On Mon, Oct 22, 2012 at 9:08 AM, Roy Stogner <roystgnr@ices.utexas.edu> wrote:

On Mon, 22 Oct 2012, Kirk, Benjamin (JSC-EG311) wrote:

To be clear, 4.5 is a proper release.

The issue is they distribute two packages - the peon version
consisting of two files (which we propose we move to) and the
developers version for Jose who want to muck with fparser.

We presently have packaged the development version - the suggestion
is instead we package the peasant version.

Ah, thanks for the clarification.  I'd rather package the development
version, even if we've no plans to muck with it, if only to make sure
we stay GPL-compatible.  Those .inc files are basically wads of
intermediate bytecode, right?

On the other hand, not breaking on OS X is definitely a higher
priority, so let's switch back to release if we must.

I like the "include both, selectable at configure-time" option best,
if it's not too much work for Cody.  In exchange I can finally get
around to reproducing/fixing that Nemesis regression he found...

Well my results are still very perplexing:

Using Mac Ports I have installed gcc45, gcc46 and gcc47 - with no other normal external packages like PETSc or MPI for this test:

trunk with gcc45:
opt - PASS
dbg - FAIL (fparser)

trunk with gcc46:
opt - PASS
dbg - FAIL (fparser)

trunk with gcc47:
opt - PASS
dbg - FAIL (fparser)

automake with gcc45:
opt - PASS
dbg - PASS

automake with gcc46:
opt - PASS
dbg - PASS

automake with gcc47:
opt - FAIL (fparser)
dbg - FAIL (fparser)