Several verifications fail in building 1.1.

[jhallum@…]

I'm building Duchamp 1.1 for Mary Putman, and it compiles fine, except that the Verifications fail on each test:

---

Running the first Duchamp test:

[This is a simple sigma-clipping search]

PGPLOT /xw: cannot connect to X server [] ERROR <Plot Moment Map> : Could not open PGPlot device /xs. Done. Comparison to standard result: Differences with Results: 42c42 < 5 J055425-241534 74.9 74.9 108.1 05:54:25.16 -24:15:34.84 1439.420 43.37 48.66 145.104 383749.031 2961.102 11.304 11.19 70 80 69 80 103 114 492 74.9 74.9 108.2 74.9 74.9 108.1 74 75 107 ---

5 J055425-241534 74.9 74.9 108.1 05:54:25.16 -24:15:34.84 1439.420 43.37 48.66 145.104 383749.000 2961.102 11.304 11.19 70 80 69 80 103 114 492 74.9 74.9 108.2 74.9 74.9 108.1 74 75 107

Differences with Log: Running the second Duchamp test:

[This uses the FDR method]

PGPLOT /xw: cannot connect to X server [] ERROR <Plot Moment Map> : Could not open PGPlot device /xs. Done. Comparison to standard result: Differences with Results: 38c38 < 1 J060912-273501 25.1 25.1 20.0 06:09:12.24 -27:35:01.09 277.618 52.66 43.29 158.296 352515.188 2924.834 11.489 11.37 19 31 20 30 14 26 484 25.1 25.1 20.1 25.1 25.1 20.0 26 25 21 ---

1 J060912-273501 25.1 25.1 20.0 06:09:12.24 -27:35:01.09 277.618 52.66 43.29 158.296 352515.219 2924.834 11.489 11.37 19 31 20 30 14 26 484 25.1 25.1 20.1 25.1 25.1 20.0 26 25 21

41c41 < 4 J060859-241502 25.1 75.1 86.1 06:08:59.90 -24:15:02.38 1148.797 44.67 51.43 158.296 353717.188 2843.094 10.445 10.33 20 30 69 81 81 93 470 25.1 75.1 86.1 25.1 75.1 86.1 26 75 86 ---

4 J060859-241502 25.1 75.1 86.1 06:08:59.90 -24:15:02.37 1148.797 44.67 51.43 158.296 353717.156 2843.094 10.445 10.33 20 30 69 81 81 93 470 25.1 75.1 86.1 25.1 75.1 86.1 26 75 86

Differences with Log:

Running the third Duchamp test:

[This spatially smoothes the cube, then searches with simple sigma-clipping]

PGPLOT /xw: cannot connect to X server [] ERROR <Plot Moment Map> : Could not open PGPlot device /xs. Done. Comparison to standard result: Differences with Results: 49c49 < 5 J055424-241517 74.9 75.0 108.0 05:54:24.61 -24:15:17.15 1438.174 55.30 52.85 211.061 686542.562 3983.107 11.304 11.19 69 82 69 81 99 115 783 75.0 75.0 108.0 74.9 75.0 108.0 74 75 107 ---

5 J055424-241517 74.9 75.0 108.0 05:54:24.61 -24:15:17.15 1438.174 55.30 52.85 211.061 686542.500 3983.107 11.304 11.19 69 82 69 81 99 115 783 75.0 75.0 108.0 74.9 75.0 108.0 74 75 107

Differences with Log: Running the fourth Duchamp test:

[This reconstructs the cube, then searches with simple sigma-clipping] [It should take a bit longer than the others.]

PGPLOT /xw: cannot connect to X server [] ERROR <Plot Moment Map> : Could not open PGPlot device /xs. Done. Comparison to standard result: Differences with Results: 47c47 < 4 J060900-241502 25.1 75.1 86.0 06:09:00.12 -24:15:02.60 1148.294 44.53 39.42 211.061 360494.312 3351.403 10.445 10.75 20 30 70 79 80 96 641 25.0 75.1 86.0 25.1 75.1 86.0 26 75 86 ---

4 J060900-241502 25.1 75.1 86.0 06:09:00.12 -24:15:02.60 1148.294 44.53 39.42 211.061 360494.344 3351.403 10.445 10.75 20 30 70 79 80 96 641 25.0 75.1 86.0 25.1 75.1 86.0 26 75 86

Differences with Log: No differences? Then everything worked as expected!

---

I have built 1.0.5 successfully on the machine without no verification errors, and I used the same INC and LIB variables for 1.0.5 and 1.1. Where is a good place to look first to figure out where these errors are creeping in? Thanks for your help.

[MatthewWhiting]

Sorry for delay in replying -- there have been problems with the ticket notification system...

This does not seem to be a problem with the Duchamp code itself -- I would say it compiled fine and should run perfectly. It looks as if there is a difference in the way the integrated flux is calculated between your machine and the one on which the comparison results were generated. It is probably a precision-related issue. I haven't seen this happen on previous versions, so it is likely something related to recent updates. I'll try and track this down, but I'll just emphasise that I don't expect there to be any problems with the Duchamp build.

It also seems that you don't have pgplot access on the terminal you are running the test on -- this is not needed for the verification scripts, so I will add the -x flag to the verification scripts.

Matt Whiting.

[MatthewWhiting]

There is a minor release (v1.1.1) out today that has an improved verification script.

There is not much substantially different to v1.1, so you should be fine using the version you've installed. The main relevant change is improved precision in calculating the integrated flux (the cause of the diffs you were seeing).

But as I said before, the version you installed appears to be working fine, it was just a precision-related issue.

Matt Whiting.