Survex 1.1.16 development snapshot uploaded

[Olly Betts]

On Mon, May 16, 2011 at 10:07:33AM +0100, David Loeffler wrote:
> It's fair enough if the current version is a bit more fascist about
> dates than 1.0 versions were; but if the errors weren't fatal, why
> weren't any output files produced? How do I force it to ignore the
> non-fatal errors and write a 3d file anyway? The full transcript
> (below) makes it clear that the actual calculations succeeded and
> hence cavern knows perfectly well where the stations are, it's just
> sulking and refusing to output them to a file!

The terminology here is that a "fatal error" is one which means that the
program simply can't continue (so something like a disk error, or
running out of memory).  An non-fatal error is only non-fatal in that
we don't have to bail out right away - there's nothing which completely
prevents further progress.

(I'm not sure this is actually a useful way to think about it - this has
raised similar questions before (i.e. "why doesn't it produce files if
the errors are non-fatal?").  Probably just dropping "non-fatal" here
would actually be clearer.)

The main reason we keep going is to show further errors, so the user can
see everything which needs fixing at once, avoiding the extremely
irritating syndrome where they get to fix just one error for each
reprocessing.

I can see a case for a bad date being a warning - the real mitigating
factor is that cavern used to be less fussy here, so there are datasets
which used to process but no longer do.

Cheers,
    Olly