[daip] g77 Dw.d specifier / UVFITS

[Eric Greisen]

Athol Kemball writes:
 > 
 > I have received some UVFITS files recently and noticed that the double 
 > precision table
 > keywords appear with exponent letter 'E' in place of 'D', so that double 
 > precision and
 > real type are not distinguished. Versions of g77 before gcc 3.5 (which has 
 > yet to be
 > released) all output Dw.d with exponent 'E' instead of 'D'. This is 
 > formally allowed in
 > the F77 standard, but not common practice with F77 compilers. This behaviour is
 > in gcc 2.95.3 and successors 3.x, but will be fixed in 3.5 I believe. This 
 > can be
 > easily filtered on output in FITTP - and might be a good precaution against E/D
 > ambiguities of this type.

The FITS standard recognizes only integer and real (floating value)
numeric keywords.  There is no implication of precision and no
diferentiation between an E and a D in the exponent - in the FITS
standard.  I suppose that FITTP and FITAB could try to write with a D
format when writing double precisions (at present E format is used
almost everywhere although the number of digits is larger for doubles
usually) since this might convey a useful implication.  But table
header keywrods in aips are all doubles now for simplicity and so the
implication is not really right in some of these cases.

Eric Greisen