lists should be reverse-compiled into targetlist index numbers, because
that's the only interpretation the parser allows for a constant in these
clauses.  (Ergo, the only way they could have gotten into the list in
the first place is to have come from the targetlist; so this should always
work.)  Per problem report from Peter E.

join clauses.  The mergejoin executor wants all the join clauses to appear
as merge quals, not as extra joinquals, for these kinds of joins.  But the
planner would consider plans in which partially-sorted input paths were
used, leading to only some of the join clauses becoming merge quals.
This is fine for inner/left joins, not fine for right/full joins.

also go there.

- Fix crash due to null string pointer in some tar files with some libs

Although it was now using the right equation, it was making bogus choices
of the precision to compute intermediate results to.  I'm not sure this
is really right even yet, but it's better than before ...

is not a complete declaration.

tag the source tree as REL7_1, renaming the old tag REL7_1_BETA like it
should have been done

generate a new Changelog from rc4, which is nice and short ... one python
interface change

tag configure as 7.1, for generating the packages ...

This is it folks ... Release 7.1 is officially here ...

so I removed it.

upgrade tags to rc4 (note lower case for packagers)

update the change log, but nothing else ...

a bit.  Regenerate INSTALL.

 This stylesheet must be present for Applixware to be happy generating
 a ToC. M$Word does not seem to care one way or the other.

mark it as RC3 ...

   (fix by Hiroki kataoka).
2) Ensure the definition of atof()(#include stdlib.h).
  (suggestion by Masaaki Sakaida).

 NetBSD/PPC from Henry Hotz, several other platform updates, and move
 QNX to the unsupported list for now.

 (which failed miserably during DST) to just compare against a time
 derived from a timestamp value.
Certainly not a direct check for a correct result, but should work
 at any time of year.

it needs to ensure that data structures attached to fmgr info records in
the trees will stick around that long, too.  Current code was crashing
on cases like datatypes with old-style I/O functions.

no room for a record on last log page.

they have children.  As it stood, -d dump mode was badly broken.

This patch is Peter E's, but I'm applying it on his behalf so it'll get
into 7.1RC3 tonight (I'm guessing Peter went to bed already).

Jason Tishler's credit, if it's broken it's my fault ...

datatype.  Perhaps someday we can figure out a way of getting reproducible
results from testing this type, but for now ...

Thanks to Tom.

starting with '$'.