bsimon@loxane.com writes:
> The thing is that I do another spi_connect inside all subqueries (query1
> ... query10) and the memory allocated inside all "subcontext" doesn't
> seems to be freed until the last call of spi_finish. (the one of the
> "big" function).
No, spi_finish frees memory used in that particular SPI invocation.
What's your evidence for thinking otherwise?
regards, tom lane