[Proj] Selftest only partly reliable, example PJ_mod_ster

[vanadovv at hetnet.nl]

Hello,
It is probably wise to give a warning against the new selftest function in the PROJ projection modules.
If the projection code is wrong, the selftest results are in absolute sense completely wrong,
but the selftest could verify the computer code as perfectly correct.
Example: PJ_mod_ster.c, as found in the repository on https://github.com/OSGeo/proj.4/tree/master/src
What is the problem?
Please refer to earlier PROJ list messages:
My message from Feb 17, 2008; 9:31am; subject: Small errors in modified stereographic
And a repeat from Frank Willett, Sep 10 2011, 10:36:42 EST, subject: Projection gs48 error?
The Lee code has an incorrect coefficient and the GS48 code has an incorrect phi0.
Assuming Snyder is correct (Map projections a working manual).
For Lee, the first selftest produces:
-25564478.9526050538, 154490848.8286255
but it should be:
-25448425.343461025, 153791681.09640324
For GS48, the first selftest produces:
-12110635.970867658000, 11668127.145744404000
but it should be:
-1923908.4465293458, 355874.65894447843
No, I cannot file a ticket, I cannot produce a patch, I cannot give a working example of command line input and output from a compiled PROJ version.
Please do whatever you want with the above information, but at the moment the PJ_mod_ster.c code is still in error.
Yours,
Oscar van Vlijmen
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/proj/attachments/20160626/88c2af69/attachment.html>