On May 8, 2006, at 1:28 PM, Dries Kimpe wrote:
> I searched the entire code for other references to 'LSFLAGS', but
> couldn't find any.
> This probably means LSFLAGS is a typo? (After all, S is next to D
> on the keyboard ;-)
> Anyway, changing LSFLAGS to LDFLAGS makes everything work again.
Yes, this is a bug -- it should be LDFLAGS instead of LSFLAGS. I'll
fix that today and get it pushed out into our release branches so
that both 1.0.3 and 1.1 will contain the fix.
> PS. Shouldn't configure warn if no io mca can be built?
> Granted, configure cannot know if another one can be loaded
> dynamically later on,
> but I did specify --enable-io-romio (which actually is the
> default), so an error/warning
> would be a appreciated. Otherwise, if you happen to miss the one-
> line output in
> the total configure output mentioning that romio will not be built,
> you end up
> with an MPI implementation returning an error code for every
> MPI_File call.
> I only found out that no MPI-IO was being included in Open MPI
> after trying to debug a non-existent
> function ;-)
We go back and forth on this one, actually. We don't want Open MPI
to fail to configure if there is no I/O component available, in part
because developer testing builds generally don't want to wait for
ROMIO to build.
However, we really *should* abort configure and warn the user if --
enable-io-romio is given and ROMIO fails to configure. I will fix
this with the LDFLAGS fix. I'll let you know when the fixes become
Open MPI developer