All,
On 2004 Oct 21 , at 21.02, Tim Jenness wrote:
> On Thu, 21 Oct 2004, Alasdair Allan wrote:
>
>>> I'm currently stuck on ccdpack, which is showing the _f__xargv
>>> problem.
>>> Very bizarre -- it doesn't appear to be actually defined anywhere!
>>> I
>>> have proved to myself that Fortran cannot possibly work on the Mac,
>>> and
>>> all evidence of it having done so in the past is a product of my
>>> diseased imagination.
>>
>> *scratch head*
>
> Brad or I have not seen this at all. And we have been using shared
> libraries
It's very weird, and it might possibly just be an artefact of the
stop-start build, which wouldn't appear if I tried it with a fresh
checkout (I'm not about to test that hypothesis). So Al, don't kill
your current build -- you never know, it may just sail gaily past
CCDPACK (mention that, if it happens).
>>> I've also discovered (building dipso) that libsnx and libncar appear
>>> to
>>> define many of the same symbols. While the Linux linker must have
>>> been
>>> happy enough with this, the OSX linker gets very cross indeed, and
>>> says
>>> many rude things.
> Note that libsnx and libncar are static libraries because they share a
> common block.
Right: I hadn't thought of that. Of course, I should be able to turn
off shared libraries for this component. In fact static (convenience)
libraries which share a common block (that is, subpar and parsecon) can
be linked into a shared library if the right Fortran juju is applied,
but that's a different solution.
> I also thought that
> multiple symbol definitions were sometimes a warning from the compiler
> (at least for shared libraries - it likes to warn about the link being
> order dependent) so can sometimes be ignored.
The OSX linker doesn't do _warnings_.
I think I can predict quite confidently where it would come down in the
smacking debate: `this hurts me more than it hurts you, you know'.
I'll get the poster into a state where I can print something tomorrow,
but you're right, David, I'm sure there are print shops in Pasadena, if
it absolutely comes to that.
I'll still try to work on the build, though. Brad will have a working
build, but it would be good if a fresh checkout worked for folk there.
Norman
--
----------------------------------------------------------------------
Norman Gray : Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk
|