Release Notes for Mariposa (alpha-1)


If any late-breaking news or installation tips become
available, this is where you'll find them.

If you have any questions or comments, please send
them to:

	mp-questions@postgres.berkeley.edu


Current open bugs:

                                   Buglist

  [Mariposa]
              bug ID: 2
              summary: Aggregate functions don't work on non-attribute
              expressions.
              description: While trying to run tpc-D query #1, which
              includes aggregates like "float8sum(L_EXTENDEDPRICE * (1.0 +
              L_TAX) * (1.0 - L_DISCOUNT))" I keep getting the error
              "lookup of type 0 failed". The three attributes are all
              float8, but the parser for some reason doesn't fill in the
              correct type for the expression.
              status: open

              bug ID: 5
              summary: Stored SQL procedures leak memory.
              description: When a query includes a stored SQL procedure, it
              leaks memory. I think this is becuase postquel_start() now
              calls FragmentRTable() and FillInFragStoreids() which
              eventually leads to palloc() calls to put new stuff in the
              range table. This memory will eventually be freed at the end
              of a transaction, but if the stored procedure is called many
              times (e.g. on every tuple in a large table) then it runs out
              of memory. I think the range table info. should be
              initialized once for stored procedures and that they should
              go through the normal brokering process, in spite of what the
              comment in postquel_start() says.
              status: open

              bug ID: 6
              summary: Optimizer always adds in cost of sorting for
              merge-joins
              description: The optimizer always adds in the cost of sorting
              the two join relations for a merge-join, even when one or
              both of them is indexed. The procedure cost_mergesort()
              should be changed to check for indexes.
              status: open

              bug ID: 13
              summary: Remote index scans are not supported.
              description: jsidell 4/16/96 The optimizer will never
              consider index scans if the underlying relation is remote,
              since index_info() only scans the local pg_index catalog.
              Either: 1) index_info() should be changed to call the name
              server, or 2) the remote site can check for available indices
              when it receives an offer (either long or short protocol - if
              long protocol, the bidder should check for indices during the
              bidding process also)
              status: open

              bug ID: 16
              summary: select on modified column causes postmaster to crash
              description: The regression test includes several queries to
              test the inheritance system; there are bugs dating from
              Postgres days, and in fact appear in Postgres95. One now
              causes Mariposa to crash. I tappears that there is a problem
              in the caching system when an inherited table has a column
              name altered. The offending query is: SELECT class, aa, a
              FROM a_star*; (line 2513 of the queries.source file in
              src/test/regress)
              status: open

              bug ID: 18
              summary: insert into a remote fragment causes sitemanager to
              crash
              description: [6/23/96, am] When running a distributed system,
              doing an insert into a remote fragment causes the sitemanager
              to crash. This may be a nameserver bug. (Needless to say,
              this is a high-priority bug.) The workaround is to do the
              insert from a session running on the remote system.
              status: open