>   _a avea (to have) + să + subjunctive (which differs from present indic.
> only in 3rd person forms):
> am să merg, ai să mergi, are să mearga, avem să mergem, aveți să mergeți,
> au să mearga

Minor correction to what Ray said: in this construction you must use
the short forms of _a avea_, which are different in the 3sg, 1pl and
am să merg, ai să mergi, *a* să meargă, *am* să mergem, *ați* să
mergeți, au să meargă.

The 3sg construction here is the source of the 3rd construction Ray
mentioned w/ "o să": the particle _a_ is pronounced _o_ in many
dialects, so "o să meargă" ( < "a să meargă") was borrowed across
dialect boundaries and generalized.

> 3.[CASUAL]
> An invariable particle _o_ followed by a să-clause:
> o să merg, o să mergi, o să mearga, o să mergem, o să mergeți, o să mearga
> If the Romanians wrote _o să_ as one word, the lasy construction would be
> identical with the modern Greek way of forming the future, i.e. an
> variable particle + subjunctive.

JS Bangs
[log in to unmask]

"I could buy you a drink
I could tell you all about it
I could tell you why I doubted
And why I still believe."
 - Pedro the Lion