[Proj] proj 4.9.1 released

sisyphus1 at optusnet.com.au sisyphus1 at optusnet.com.au
Wed Mar 11 07:12:58 EST 2015


-----Original Message----- 
From: Howard Butler
Sent: Wednesday, March 11, 2015 6:44 AM
To: PROJ.4 and general Projections Discussions
Subject: [Proj] proj 4.9.1 released

> All,
>
> proj 4.9.1 is released as approved by the MetaCRS list. It cumulates a 
> number of bug fixes and small updates, and it went through five release 
> candidates before being declared final.
>
> http://download.osgeo.org/proj/proj-4.9.1.tar.gz

I notice that projects.h is being installed again by 'make install'. Was 
that intended ?
If so, it's a welcome re-introduction.
IIRC, recent builds of proj stopped installing projects.h because it was 
considered non-public API.
No big deal ... just curious. I kept on (manually) installing projects.h, 
anyway.

I've just built 4.9.1 on native MS Windows using a variety of mingw ports of 
32-bit and 64-bit gcc compilers - using autotools in the msys shell.
(I hope that autotools support is here to stay for a while yet - that there 
are no plans to soon *replace* it with cmake.)
It's the most trouble-free Windows build of proj that I've ever struck - 
configure and make worked flawlessly. (Not that the build usually threw up 
serious problems, anyway - but kudos to all involved.)

There's a (what strikes me as probably minor) problem with 'make test' in 
that one of the tests fail:

doing tests into file tv_out, please wait
Rel. 4.9.1, 04 March 2015
<cs2cs.exe>: while processing file: <stdin>, line 1
pj_transform(): invalid x or y
Rel. 4.9.1, 04 March 2015
<cs2cs.exe>: while processing file: <stdin>, line 2
pj_transform(): acos/asin: |arg| >1.+1e-14
diff tv_out with tv_out.dist
310c310
< -140.100000 -87.000000 
   987122.418330275450     -14429896.539530911000 0.000000000000
---
> -140.100000 -87.000000 
>    987122.418330275454     -14429896.539530910552 0.000000000000

Yes - shamefully, I hadn't tested any of the release candidates.
I note that if the results are rounded to 17 decimal digits then there's no 
discrepancy.
Assuming there's some interest in sorting out this failure for the future, 
what other info do I need to provide ?
What are some of the things I can try here to determine what has gone wrong 
?

I built proj using 32-bit ports of gcc-4.7.0 & gcc-4.8.2, and 64-bit ports 
of gcc-4.7.0, gcc-4.8.2 & gcc-4.9.2. In all cases the test failure was the 
same.

Cheers,
Rob




More information about the Proj mailing list