Christoph Berg <cb@df7cb.de> writes:
> Re: Tom Lane 2015-01-10 <22432.1420915326@sss.pgh.pa.us>
>> So what I propose we do with this is patch HEAD and 9.4 only.
>> We need to do *something* in 9.4 to address Christoph's complaint, and
>> that branch is new enough that we can probably get away with changing
>> officially-unsupported APIs. The lack of other field complaints makes
>> me okay with not trying to fix these issues further back.
> The problem isn't present in 9.3 and earlier (at least with
> postfix-pgsql), so there's no need to go back further.
Right, the specific issue you're complaining of is new in 9.4. The
general issue that failures in /etc/passwd lookups aren't well reported
has been there more or less forever, but your immediate point is that no
such failure should be reported to the user at all if he's supplied a
login name to use.
It looks like the bad handling of Windows' GetUserName() failures has
been there a long time too, but given the lack of field reports I don't
feel a need to back-patch that very far either.
regards, tom lane
