Andrew Ross wrote:
On 2007-10-02 12:48+0100 Andrew Ross wrote:
F95 example 28 now compiles and works. Visually I get identical
results to the other languages. ifort produces identical files.
gfortran 4.1 does not quite produce identical files. When plotting the
string "The future of our civilization depends on software freedom."
all the spaces get optimised away and nothing is plotted. For other
languages (e.g. C, f77) a space is plotted. Visually this makes no
difference, but it does mean the files are not identical. I assume
this is something to do with the implementation of transfer in
gfortran. It would be interesting to see if other people suffer the
I did uncover a bug in sfstubs.f90 which meant that plbox3 did not
work as expected with f95. This showed up in example 11 and 28. Now
Thanks for your recent f95 work. I also ran into the problem of blanks
turning into no-ops for the f77 bindings and example 28. I fixed it with an
islen change so that trailing blank trimming occurs as before unless the
string is all blanks in which case a null-terminated string with a single
blank is produced rather than the empty null-terminated string that was
I have just made two commits to bindings/f95 to make those bindings
compatible with my recent f77 changes (including the islen change). I don't
have access to Fortran 9x so please check that these changes (a) don't break
anything and (b) solve the problem of blanks turning into no-ops for example
28 that you have just described.
With my recent bindings changes do all the f95 examples now produce
identical results to the C examples? For the f77 examples I had to do some
blank tweaking (e.g., change some ' ' strings to '') to make that happen.
I can confirm this now produces identical results with ifort and with
that is excellent news. So the new example has caused some improvements
F90 bindings as well as an extension of the bindings with the 3D text