Bidirectional bond type specification required?

I has specified my bond order in following way:
bondtype C3 O2 double

To my surprise in the bond dump list I had found (bond order ignored):
connect 2 1 0 0 0

So I had tested following:

bondtype C3 O2 double
bondtype C2 O3 double

And I have finally get:
connect 2 1 0 0 0 double

Is this a future or a bug ? I believe the direction of bond have no influence on its type …
I can supply full .gin for reproduction. I had simplified a 100 atom system to only 2 atoms to reproduce the issue …

In addition it look like the bond defined by the bondtype are completely ignored during force field assignment (tested with Dreading) … They are listed as e.g. following in the restart file:

connect      8     5  0  0  0 resonant 
connect      9     8  0  0  0 resonant

But according manual the definition should be:

connect 5 8 resonant  0 0 0
connect 5 20 resonant  0 0 0

So what is correct ? And are the bondtype used for FF forces assignment at all (torsion angles in Dreading case) ?

Did you read the post on 21st January under " Wrong bonding in output cif file" where this was discussed? The follow up question is whether you acted it upon it and updated your version of GULP to the latest version?

I had updated recently to 6.2.3 and I belived this is the latestt version. I had overlooked you speak about 6.3 … Thank you …