Can someone explain why postgres exhibits the following behaviour . dsn has use declare fetch on. Application sets auto
commiton and proceeds to prepare, describe , bind and execute a query. While execute is running it calls sqlcancel and
willsubsequently receive
current transaction is aborted, commands ignored until end of transaction block when it attempts to describe the number
ofcolumns on a new statement it prepared on the same connection.
Why can it not continue per other vendors. Why does it do this with sqlcancel yet allows similar sequences where a
preparemay have failed or exception is thrown by execute (ie. Div by zero etc).
There seems to be no documentation saying that the handling of sqlcancel is by design and why. Why is this not a design
issuewhich still appears to exist in current 9.x releases.
Seems like a bug to me but many posts found via Google do not appear to clarify design intent, why so and why it had
notbeen changed.