Thread

7.1beta6 on NetBSD-1.5/i386 fails geometry test in 'make check'

Giles Lean <giles@nemeton.com.au> — 2001-03-22T09:11:33Z

[ I sent a similar message with less information to -bugs earlier, and
  it got held for moderation.  The moderator might like to kill that
  one ... ]

Hi,

PostgreSQL 7.1beta6 built on NetBSD-1.5/i386 fails the polygon test in
'make check'.  The following change to resultmap makes all tests
(other than random :-) pass on this platform.

This change isn't completely correct:

1. NetBSD-1.4.2/i386 wants geometry-positive-zeros, not *-bsd

   Perhaps 1.4.X/i386 and earlier should be presumed to use the
   current form, and 1.5 and later the -bsd one?  Any floating point
   gurus know what the changes are in NetBSD that caused this?

   (1.4.2/i386 fails several other tests at the moment.)
   
2. I don't know what non-alpha, non-i386 ports want at either
   1.4.X or 1.5.

*** src/test/regress/resultmap-dist	Sat Feb 24 05:12:36 2001
--- src/test/regress/resultmap	Thu Mar 22 19:09:30 2001
***************
*** 17,22 ****
--- 17,23 ----
  geometry/.*-openbsd=geometry-positive-zeros-bsd
  geometry/.*-irix6=geometry-irix
  geometry/.*-netbsd=geometry-positive-zeros
+ geometry/i386.*-netbsd=geometry-positive-zeros-bsd
  geometry/.*-sysv5uw7.*:cc=geometry-uw7-cc
  geometry/.*-sysv5uw7.*:gcc=geometry-uw7-gcc
  geometry/alpha.*-dec-osf=geometry-alpha-precision

Regards,

Giles