Regression on buffer tests on berrie64

[robe]

Berrie64 has been failing for a while. I suspect because of the new buffer logic in GEOS.

​https://debbie.postgis.net/view/PostGIS/job/PostGIS_Worker_Run/label=berrie64/3070/consoleFull

11:15:39  ./regress/core/regress_buffer_params .. failed (diff expected obtained: /tmp/pgis_reg/test_46_diff)
11:15:39 -----------------------------------------------------------------------------
11:15:39 --- ./regress/core/regress_buffer_params_expected	2022-01-11 08:04:48.687916768 -0800
11:15:39 +++ /tmp/pgis_reg/test_46_out	2022-01-11 08:15:39.769355510 -0800
11:15:39 @@ -3,13 +3,13 @@
11:15:39  line quadsegs=2 endcap=flat|POLYGON((10 2,10 -2,0 -2,0 2,10 2))
11:15:39  line quadsegs=2 endcap=butt|POLYGON((10 2,10 -2,0 -2,0 2,10 2))
11:15:39  line quadsegs=2 endcap=square|POLYGON((10 2,12 2,12 -2,0 -2,-2 -2,-2 2,10 2))
11:15:39 -line join=mitre mitre_limit=1.0 side=both|POLYGON((148.12357 161.60116,159.5301 156.87643,160 50,159.80785 48.0491,159.2388 46.17317,158.3147 44.4443,157.07107 42.92893,155.5557 41.6853,153.82683 40.7612,151.9509 40.19215,150 40,148.0491 40.19215,146.17317 40.7612,144.4443 41.6853,142.92893 42.92893,141.6853 44.4443,140.7612 46.17317,140.19215 48.0491,140 50,140 125.85786,57.07107 42.92893,55.5557 41.6853,53.82683 40.7612,51.9509 40.19215,50 40,48.0491 40.19215,46.17317 40.7612,44.4443 41.6853,42.92893 42.92893,41.6853 44.4443,40.7612 46.17317,40.19215 48.0491,40 50,40.19215 51.9509,40.7612 53.82683,41.6853 55.5557,42.92893 57.07107,148.12357 161.60116))
11:15:39 +line join=mitre mitre_limit=1.0 side=both|POLYGON((147.65367 161.7958,160 156.68179,160 50,159.80785 48.0491,159.2388 46.17317,158.3147 44.4443,157.07107 42.92893,155.5557 41.6853,153.82683 40.7612,151.9509 40.19215,150 40,148.0491 40.19215,146.17317 40.7612,144.4443 41.6853,142.92893 42.92893,141.6853 44.4443,140.7612 46.17317,140.19215 48.0491,140 50,140 125.85786,57.07107 42.92893,55.5557 41.6853,53.82683 40.7612,51.9509 40.19215,50 40,48.0491 40.19215,46.17317 40.7612,44.4443 41.6853,42.92893 42.92893,41.6853 44.4443,40.7612 46.17317,40.19215 48.0491,40 50,40.19215 51.9509,40.7612 53.82683,41.6853 55.5557,42.92893 57.07107,147.65367 161.7958))
11:15:39  line side=left|POLYGON((150 50,150 150,50 50,42.92893 57.07107,142.92893 157.07107,144.4443 158.3147,146.17317 159.2388,148.0491 159.80785,150 160,151.9509 159.80785,153.82683 159.2388,155.5557 158.3147,157.07107 157.07107,158.3147 155.5557,159.2388 153.82683,159.80785 151.9509,160 150,160 50,150 50))
11:15:39  line side=right|POLYGON((50 50,150 150,150 50,140 50,140 125.85786,57.07107 42.92893,50 50))
11:15:39  line side=left join=mitre|POLYGON((150 50,150 150,50 50,42.92893 57.07107,160 174.14214,160 50,150 50))
11:15:39  poly quadsegs=2 join=round|POLYGON((-2 0,-2 10,-1.41421 11.41421,0 12,10 12,11.41421 11.41421,12 10,12 0,11.41421 -1.41421,10 -2,0 -2,-1.41421 -1.41421,-2 0))
11:15:39  poly quadsegs=2 join=bevel|POLYGON((-2 0,-2 10,0 12,10 12,12 10,12 0,10 -2,0 -2,-2 0))
11:15:39  poly quadsegs=2 join=mitre|POLYGON((-2 -2,-2 12,12 12,12 -2,-2 -2))
11:15:39 -poly quadsegs=2 join=mitre mitre_limit=1|POLYGON((-1.82843 -1,-1.82843 11,-1 11.82843,11 11.82843,11.82843 11,11.82843 -1,11 -1.82843,-1 -1.82843,-1.82843 -1))
11:15:39 -poly quadsegs=2 join=miter miter_limit=1|POLYGON((-1.82843 -1,-1.82843 11,-1 11.82843,11 11.82843,11.82843 11,11.82843 -1,11 -1.82843,-1 -1.82843,-1.82843 -1))
11:15:39 +poly quadsegs=2 join=mitre mitre_limit=1|POLYGON((-2 -0.82843,-2 10.82843,-0.82843 12,10.82843 12,12 10.82843,12 -0.82843,10.82843 -2,-0.82843 -2,-2 -0.82843))
11:15:39 +poly quadsegs=2 join=miter miter_limit=1|POLYGON((-2 -0.82843,-2 10.82843,-0.82843 12,10.82843 12,12 10.82843,12 -0.82843,10.82843 -2,-0.82843 -2,-2 -0.82843))
11:15:39  poly boundary rhr side=left|POLYGON((20 20,10 20,10 50,50 50,50 10,20 10,20 20),(20 20,40 20,40 40,20 40,20 20))
11:15:39 -----------------------------------------------------------------------------

[robe]

She's been failing since Jan 10th 2022.

[robe]

I see both debbie and winnie are failing too with the same complaint.

[pramsey]

I'd say just strip that test out. The exact shape of mitre joins will be slightly different now, because they were slightly broken before. So either the tests need to be GEOS-version tied or just dropped.

[bowguy]

Just curious why the 64 bit version failed and not the 32 bit. Tolerance?

[pramsey]

Probably not actually pulling the latest GEOS. The mitre change went in literally yesterday.

[robe]

Okay I checked berrie and she too is failing on that test but showing a green. Looks like maybe because she goes on to run the garden tests and her garden tests succeed, that is the green that is shown.

So I guess I got to put in an error for that so she stops on failure and doesn't go on to run the garden tests.

[strk]

I think it's better to have two expected tests results, we did this in the past.

[strk]

Yep: here's how it was done with Topology in 2.0: ​https://github.com/postgis/postgis/blob/stable-2.0/topology/test/Makefile.in#L71

[strk]

The test is NOT failing for me on 64bit Ubuntu/Linux machine using the current GEOS from main branch

[strk]

Was the -ffloat-store switch recently removed from GEOS build ?

[strk]

PostgreSQL 12.7 (Ubuntu 12.7-0ubuntu0.20.10.1) on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 10.2.0-13ubuntu1) 10.2.0, 64-bit
  Postgis 3.3.0dev - (3.2.0-196-g42fbf799f) - 2022-01-11 20:46:48
  scripts 3.3.0dev 3.2.0-196-g42fbf799f
  GEOS: 3.11.0dev-CAPI-1.16.0
  PROJ: 7.2.1

Running tests

 regress/core/regress_buffer_params .. ok in 54 ms
 uninstall .. ok (4724)

Run tests: 2
Failed: 0

[Regina Obe <lr@…>]

In fbf8a51/git:

Take out buffer mitre tests
that yield different results between GEOS 3.11 and GEOS < 3.11
References #5049 for PostGIS 3.2.0

[Regina Obe <lr@…>]

In c585d7c0/git:

Take out buffer mitre tests
that yield different results between GEOS 3.11 and GEOS < 3.11
Closes #5049 for PostGIS 3.2.1