That answer, of course, brings me to my counter example where it didn't seem to work that way. If I can illustrate, I will submit to support. If only in CVF, can I still submit (though I would guess it won't be fixed)?
I'd like to see a counter-example. There's really no point to submitting to CVF support... As you know, nothing will come of that. But if I can reproduce it in Intel Fortran, that will be useful. I will say that I use NAMELIST a lot and have never seen any problems with the order or omission of variables.
Support has validated this as an issue. "Reproduced and escalated". I guess my work around for CVF is to rewind after every read (have not verified that this will work, however).
After quite a while in the support issue area, including being "validated", I'm told this is an expected result of using namelists -- that the order of the namelist reads must correspond to the namelist entries in the file UNLESS you (like I have) choose to rewind after every read -- thus assuring yourself that users may put those namelists in in any order they choose.
Fields within the name list do not apparently matter as to order -- you can continue to have them in any order.
This is my fault. I misread your original post. I thought you were asking about the order of variables WITHIN a nameslist, and not the groups themselves. Indeed, the file is read sequentially and there is no backtracking unless you do so with a REWIND. My apologies for wasting your time with misdirection.
