I wrote: > I still dare to doubt whether you've tested this, because AFAICS > the operand numbering is wrong. The "r"(lock) operand is number 3 > given these operand declarations, not number 2.
Oh, my apologies, scratch that. Evidently I put in the "+m"(*lock) operand and confused myself about what was what. I still think the form I proposed is better style though. regards, tom lane