Am 01.11.2016 um 15:14 schrieb Fabien COELHO:
Thus, when returning with an error, if conn->pgpassfile was set and a
password was necessary, we must have tried that pgpassfile, so i got
rid of the field "dot_pgpass_used"
No, you should not have done that, because it changes a feature which
was to warn *only* when the password was coming from file.
The warning is wrong, the password was typed directly, not retrieved
from a file. The "dot_pgpass_used" boolean is still required to avoid
that.
I see... I was too focused on looking for things to declutter, that I
missed that case. I'll address that in the next revision.