Bruce Momjian <maillist@candle.pha.pa.us> writes:
> I recommend you check out what is currently done properly, and fix the
> ones that are incorrect.
Well, yes. The question was how to tell which is which :-)
> I can imagine some cases where you would want to get a lock and keep it
> until the end of the transaction, and other times when you would want to
> release it before transaction end.
I guess I'm not understanding something. How can it ever be correct
practice to release a lock before transaction end? For example, if I
write some changes in a table, and then release the lock, wouldn't that
allow other backends to see the not-yet-committed changes? What if I
then abort my transaction? Now the other backends have acted on
information they should never have seen at all.
Releasing a read lock strikes me as just as dangerous but for more
subtle reasons --- once you have read a table, what you have read
ought to look the same until the end of your transaction.
Since there is an unset-write-lock function, I assume it must have
valid uses, but I don't see what they are.
>> Is there a bug here, or is there some special definition of user access
>> to a system table that means the select isn't acquiring a read lock?
>> Selects and updates on ordinary user tables seem to interlock fine...
> Select certainly should be locking. Something is wrong, but I am not
> sure what. If you want me to check into it, let me know.
Please. (Note that I saw this with my revised version of async.c;
I believe you will see the same behavior with the currently-checked-in
code, but do not have the time to rebuild that version to make sure.)
regards, tom lane