Regress errors: ReprojectPts, ReprojectPtsGeog, tickets

[aperi2007]

Hi, compiling the r6943 on a debian lenny 32bit with postgres 9.0.2 I see these errors:

loader/ReprojectPts….. failed (wkt test: diff expected obtained: /tmp/pgis_reg_8047/test_20_diff)

loader/ReprojectPtsGeog….. failed (wkt test: diff expected obtained: /tmp/pgis_reg_8047/test_21_diff)

tickets. failed (diff expected obtained: /tmp/pgis_reg_8047/test_58_diff)

Attach the logs for pgis_reg_8047.

[aperi2007]

logs of errors

[strk]

Hudson is reporting the same problems (dunno why _to_me_ actually, but that's another story)

[jadams]

Can you please attach the contents of /tmp/pgis_reg_8047/test_20_diff and /tmp/pgis_reg_8047/test_21_diff?

Also is there a way I can see the hudson results? I added these tests (so it's probably my fault they're failing) but I didn't realize hudson was complaining.

[aperi2007]

it is add to the ticket.

[strk]

@jadams ​http://office.refractions.net:1500/job/PostGIS-trunk-regression/

[robe]

Those just look like rounding errors. Don't we have some logic in regress to truncate end decimals?

BTW: tickets is failing for me too on MingW - but I suspect its somethin else that changed perhaps something Paul changed. I was blaming Olivier but I don't think he committed anything during the sprint.

[jadams]

I agree they look like rounding errors. I had increased the precision of the output, which improved but clearly didn't fix the problem.

There is not anything in universal regress to truncate decimals though it is a good idea.

[robe]

ST_SnapToGrid or did you try that already. I think that's how we've dealt with these in the past.

[strk]

round(x * 1eY) / 1eY

with Y being number of decimal digits

[robe]

I put in an ST_SnapToGrid call and it still works for me. strk and whoever was having problems can you test it and see if it passes for you

[nicklas]

Regina, I think it was mostly Hudson complaining. Shortly after your commit Hudson sent a message that it was satisfied.

That is quite strange, isn't it? Why did the ticket issue disappear? And then come back when Pierre removed some issues in raster.

Strange!

/Nicklas

[robe]

there are 2 jobs one is the build and the other is the regress (took me a while to figure that out since I noticed the strange thing). The regress is the one that has been failing usually. The build is usually successful — that's why the site docs and tar file on site is up to date.

[aperi2007]

Hi, I tested again using geos 3.3.0 trunk r3275 and postgis trunk r6978 on a debian lenny 32bit with postgres 9.0.1.

Some error was disappeared. but still having the ticket error.

tickets. failed (diff expected obtained: /tmp/pgis_reg_21420/test_59_diff)

I put the logs in attach to this ticket.

[aperi2007]

log for ticket error

[robe]

well Hudson seems to be happy with my changes so I'll take that as a green light to close this one out. That just leaves — #869 to make the regress completely pass.