Robert Haas <robertmhaas@gmail.com> writes:
> Ugh, yes. Sorry, I should have checked this more carefully before
> commit. I mentioned the problem in a previous review and failed to
> notice that it hadn't been fixed. Are you taking care of it at this
> point or should I keep at it?
I'm about to push a fix that removes the crashes (or at least the ones
I see on dromedary), but there is still a problem, which is that the
expected output seems inherently platform-dependent:
*** /Users/tgl/pgsql/contrib/pageinspect/expected/hash.out Thu Feb 2 21:30:54 2017
--- /Users/tgl/pgsql/contrib/pageinspect/results/hash.out Thu Feb 2 22:55:43 2017
***************
*** 52,58 ****
magic | 105121344
version | 2
ntuples | 1
! ffactor | 307
bsize | 8152
bmsize | 4096
bmshift | 15
--- 52,58 ----
magic | 105121344
version | 2
ntuples | 1
! ffactor | 384
bsize | 8152
bmsize | 4096
bmshift | 15
***************
*** 107,113 ****
live_items | 1
dead_items | 0
page_size | 8192
! free_size | 8128
hasho_prevblkno | 4294967295
hasho_nextblkno | 4294967295
hasho_bucket | 2
--- 107,113 ----
live_items | 1
dead_items | 0
page_size | 8192
! free_size | 8132
hasho_prevblkno | 4294967295
hasho_nextblkno | 4294967295
hasho_bucket | 2
======================================================================
I think probably both of those are unavoidable 32-bit v 64-bit
differences due to available space on a page changing with MAXALIGN.
What do you want to do about those?
regards, tom lane