lwgeom_regress test fails on i386

[mwanner]

Any build of postgis-2.0.3 or source:/branches/2.0@11591​ on i386 (and within a chroot) fails 'make check' for me.

2.0.3 only fails on lwgeom_regress, r11591 shows the additional 'tickets' failure. None of these show up on 64bit builds (i.e. amd64).

Everything below is from a build of r11591: ./configure says:

  PostGIS is now configured for i486-pc-linux-gnu

 -------------- Compiler Info ------------- 
  C compiler:           i486-linux-gnu-gcc -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security
  C++ compiler:         i486-linux-gnu-g++ -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security
  SQL preprocessor:     /usr/bin/cpp -traditional-cpp -P

 -------------- Dependencies -------------- 
  GEOS config:          /usr/bin/geos-config
  GEOS version:         3.3.3
  GDAL config:          /usr/bin/gdal-config
  GDAL version:         1.9.0
  PostgreSQL config:    /usr/bin/pg_config
  PostgreSQL version:   PostgreSQL 9.1.9
  PROJ4 version:        47
  Libxml2 config:       /usr/bin/xml2-config
  Libxml2 version:      2.9.1
  JSON-C support:       yes
  PostGIS debug level:  0
  Perl:                 /usr/bin/perl

 --------------- Extensions --------------- 
  PostGIS Raster:       enabled
  PostGIS Topology:     enabled

 -------- Documentation Generation -------- 
  xsltproc:             /usr/bin/xsltproc
  xsl style sheets:     /usr/share/xml/docbook/stylesheet/nwalsh
  dblatex:              /usr/bin/dblatex
  convert:              /usr/bin/convert
  mathml2.dtd:          http://www.w3.org/Math/DTD/mathml2/mathml2.dtd

Failing tests:

lwgeom_regress .. failed (diff expected obtained: /tmp/pgis_reg/test_27_diff)
tickets .. failed (diff expected obtained: /tmp/pgis_reg/test_66_diff)

$ cat /tmp/pgis_reg/test_27_diff
--- lwgeom_regress_expected	2013-04-10 15:34:25.000000000 +0000
+++ /tmp/pgis_reg/test_27_out	2013-07-19 07:48:22.998406489 +0000
@@ -1,5 +1,5 @@
 BOX(0 0.1,11 12)
-BOX3D(0 0.1 -55,11 12 12)
+BOX3D(0 0.1 -55,11 12 4.75497462555748e+104)
 11184
 15824
 20464

$ cat /tmp/pgis_reg/test_66_diff
--- tickets_expected	2013-07-19 07:16:53.000000000 +0000
+++ /tmp/pgis_reg/test_66_out	2013-07-19 07:48:32.998407050 +0000
@@ -233,4 +233,4 @@
 #2110.1|f
 #2110.2|t
 #2110.3|t
-#2307|MULTIPOLYGON(((-41.1932 -7.3257,-41.1616 -7.3257,-41.1569 -7.3257,-41.1569 -7.3483,-41.1932 -7.3483,-41.1932 -7.3257),(-41.1616 -7.3257,-41.1879 -7.3257,-41.1879 -7.3425,-41.1616 -7.3425,-41.1616 -7.3257)))
+#2307|MULTIPOLYGON(((-41.1932 -7.3257,-41.1616 -7.3257,-41.1569 -7.3257,-41.1569 -7.3483,-41.1932 -7.3483,-41.1932 -7.3257),(5.78163017139907e+194 4.36414105401949e+254,27883.4492 -4.2942574203874e+301,6.46654932641944e+118 7.54575043065372e+107,9.90794679509237e+117 6.46734808040511e+118,-41.1616 -7.3257)))

Also note that I initially suspected my own patches for Debian to cause this. However, all of the above applies to the unpatched release or svn checkout as well.

[robe]

FWIW: The first error is one I've seen before under our windows mingw64-w32 build bot a while ago. For the BOX3D case I ended up just changing the offending test as described in #2020 (though your failure might be different reason). Because I thought it was an unfair test and the last coordinate is not predictable from the code.

Your second error test_66_diff I suspect your regress is hooking up with an ancient lwgeom.so you have installed in system. The reason I say that , is the second test is a fairly new test put in to test that particular issue, so if you were running an out of date lwgeom, you'd fail that test.

[robe]

Backported the test change from 2.1 to 2.0 at r11704

See if that reduces your failures down to 1.

[robe]

Haven't heard anything. I'm going to close this ticket out. Feel free to reopen if you still have issues.