Fix to split code for single large record at the end of a page.

cast alloc's to shut SunOS up

first pass at making DB use basic integral types (the Alpha port)

4.4BSD snapshot (revision 8.1); add 1993 to copyright

recno bug -- if split to a new page on index 0 (a sorted split) was
changing a random value as the new number of records on the left page

break BTF flags up into btree/recno (B_, R_) categories

more lint

fix some types for the Sun compiler

bcmp/bcopy/bzero -> memcmp/memmove/memset

index -> indx_t

move db.h to the end of the includes, make compat.h work better

if large key/data pairs, can split and still not have enough room for
the new key. Pass the length of the new key into the split routines so the
calculation is correct. Remove prefix compression w

if large key/data pairs, can split and still not have enough room for
the new key. Pass the length of the new key into the split routines so the
calculation is correct. Remove prefix compression when BINTERNAL pages are
being split -- it's not possible (Bayer's paper agrees). Most of the actual
split code was reworked, including cursor adjustment.

show more ...

Put new key/data pair into split page before fixing up the root page
if the root page split. The old method, of guaranteeing that the split wasn't
on the leftmost key of the right page doesn't work,

Put new key/data pair into split page before fixing up the root page
if the root page split. The old method, of guaranteeing that the split wasn't
on the leftmost key of the right page doesn't work, there are cases where the
split HAS to be that key. Example is three keys, A, B, D, each 1/3 of a page
and an insert of a key C which is 1/2 of a page.

show more ...

abort() if unrecognized switch value; initialize nksize (thanks GCC!)

implement free page reuse

once half < nbytes, don't decrement further, can wrap around

bytes needed calculation for BLEAF parent page was wrong -- included
data bytes as well. This caused a later error, not positive that it's fixed.
Also, page was being allowed to split on "skipped" o

bytes needed calculation for BLEAF parent page was wrong -- included
data bytes as well. This caused a later error, not positive that it's fixed.
Also, page was being allowed to split on "skipped" offset, which caused root
split to drop core

show more ...

change bt_nrecs and set BTF_METADIRTY, BTF_MODIFIED elsewhwere; fix bug
using rchild->flags instead of bl->flags to get bigkey flag; increment parent
pages in search routines, so just clear stack; fi

change bt_nrecs and set BTF_METADIRTY, BTF_MODIFIED elsewhwere; fix bug
using rchild->flags instead of bl->flags to get bigkey flag; increment parent
pages in search routines, so just clear stack; fix cursor repositioning code
in routine doing the physical split

show more ...

Rev #2. Redo the btree code add recno's and checkpoint

ANSI prototypes

date and time created 91/01/23 08:12:46 by mao