From: Arlindo da S. <da...@al...> - 2010-07-11 14:59:31
|
All, I'm making progress towards the 2.0.a8 upgrade. 1) The supplibs have been upgraded to version 2.2.0. To get the fresh sources and build, % gacvs co -d supplibs-2.2.0 -P supplibs % cd supplibs-2.2.0/src % make install So far I've tested this on Mac OS X 10.5.8, RHEL 5 (x86_64) and RHEL4 (i686). 2) The Grads codebase has been merged with COLA's. Since COLA now has a single executable, I've replaced our previous build with COLA's, making small adjustments for supporting the extensions and the extra-platforms we support. I am afraid we have lost the integration with pkg-config, something I could restore later. However, the close we track COLA's codebase the less the maintenance costs on our end. To check it out and build, % gacvs co -d Grads-2.0.a8 -P Grads % ln -s /path/to/your/supplibs/arch supplibs % cd Grads-2.0.a8 % configure % make % make check It passes all tests on Mac OS X 10.5.8 (Leo), x86_64 and i686. I'll try Mac OS X 10.6 (Snow Leo) and Windows once I have robust builds on these 3 platforms. 3) For testing the extensions: % make binstall (put opengrads/Contents in your path) % cd extensions % make check It passes all 93 tests on Darwin and x86_64. However, it fails a single test on i686: the creation of grads_grib files with the LATS extension (all other formats work). I love these tests... Mike, To reproduce the problem, simply try (using lats4d.sh under opengrads/Contents): % lats4d.sh -i model.nc -o junk -format grads_grib This will complete successfully, but when you try to read the file all variables will be UNDEF. I tried downgrading the LATS extension to the 2.0.a7.oga.3 version (that is, before your recent mods) but the problem persisted. The previous 2.0.a7.oga.3 binaries work with the current LATS extension. So it may be related to some internal grads changes, probably related to gribmap, but I am not sure. Could you take a look at it? Thanks, Arlindo -- Arlindo da Silva da...@al... |
From: Arlindo da S. <da...@al...> - 2010-07-12 16:19:58
|
On Sun, Jul 11, 2010 at 1:14 PM, Arlindo da Silva <da...@al...>wrote: > All, > > I think I narrowed down the problem to a possible bug in the gribmap > utility in grads 2.0.a8. The LATS extension I mentioned earlier writes out a > .grb and .ctl file, using the "gribmap" utility to create the index file > (thru a system() call). When I manually recreated the gribmap file (.gmp) by > hand using the gribmap fom 2.0.a7 it all worked fine. Therefore I suspect > there may be an issue with the 2.0.a8 gribmap utility on the i686platform, at least with the version that I built. I tried running the i686 > binaries running by COLA but it does not work on my i686 platform: > > gribmap -0 -i junk.ctl > /home/dasilva/gribmap: error while loading shared libraries: requires glibc > 2.5 or later dynamic linker > > > I am running: > > gcc (GCC) 3.4.6 20060404 (Red Hat 3.4.6-11) > Linux voda.gsfc.nasa.gov 2.6.9-89.0.23.ELsmp #1 SMP Fri Mar 5 17:56:21 EST > 2010 i686 i686 i386 GNU/Linux > > > I am attaching the .grb and .ctl files in question. To reproduce the > problem, just run > > % gribmap -0 -i junk.ctl > > and see if you can open the resulting dataset in grads. When using gribmap > in a8 I get all undefs. > > Jennifer, Hoop: > > Can you confirm whether we have an issue with 2.0.a8 gribmap on i686? > As I said earlier, I have no problems on x86_64 and Mac OS X 10.5.8. This > appears to be 32-bit Linux specific. > > Many thanks to Hoop and Jennifer for verifying that their i686 build of *gribmap* worked OK on the sample grib file I originally attached to this message; only my build was broken. I am happy to report that I have now produced a successful build of 2.0.a8.oga.1 on i686. The problem I was experiencing could be traced to the fact that I was building with gcc 3.x and glibc<2.5, a combination that up to now had produced very portable builds. For some reason that I decided not to dig into, the new gribmap appears to require gcc 4.x and glibc>=2.5. I have also made a 64-bit build on Mac OS X 10.6 (Snow Leopard). It passes all basic grads tests, but there seem to be issues with a couple of extensions that I need to look into. I'll see what I can do, but it won't hold up release as the Leopard build works just fine on Snow Leopard. This required HDF-4 to be upgraded to 4.2.5, something that I am yet to check in to the baselibs. After I do this, I'll recheck the build on the other platforms and upload the supplibs to sf.net. I'll then attempt a Windows build. Jennifer, Is the doc/ directory in your 2.0.a8 source tarball up to date? I'd like to update the documentation we ship with the bundle (for off-line use). Cheers! Arlindo -- Arlindo da Silva da...@al... |
From: Jennifer A. <jm...@co...> - 2010-07-12 18:37:20
|
On Jul 12, 2010, at 12:19 PM, Arlindo da Silva wrote: > Jennifer, > > Is the doc/ directory in your 2.0.a8 source tarball up to date? > I'd like to update the documentation we ship with the bundle (for > off-line use). I had updated the a8 code with new doc files but forgot to commit new versions of existing files. Here is a link to a tarball (it will unpack as ./doc/* ) that has all the documentation web pages, current as of today. ftp://iges.org/pub/jma/dev20/grads-doc-20100712.tar.gz --Jennifer |
From: Arlindo da S. <da...@al...> - 2010-07-13 01:54:47
|
All, I have posted binaries tarballs for supplibs 2.2.0 and grads 2.0.a8.oga.1 on sf.net: https://sourceforge.net/downloads/opengrads/grads2/2.0.a8.oga.1/ https://sourceforge.net/downloads/opengrads/supplibs/2.2.0/ If someone would like to give it a try before I make a general announcement I'd really appreciate it. I still need to finish the Snow Leopard and Windows builds. I'll post the final sources when I finish all the other builds. For now, use the checkout instructions below to build from sources. Cheers! Arlindo On Sun, Jul 11, 2010 at 10:58 AM, Arlindo da Silva <da...@al...>wrote: > All, > > I'm making progress towards the 2.0.a8 upgrade. > > 1) The supplibs have been upgraded to version 2.2.0. To get the fresh > sources and build, > > % gacvs co -d supplibs-2.2.0 -P supplibs > % cd supplibs-2.2.0/src > % make install > > So far I've tested this on Mac OS X 10.5.8, RHEL 5 (x86_64) and RHEL4 > (i686). > > 2) The Grads codebase has been merged with COLA's. Since COLA now has a > single executable, I've replaced our previous build with COLA's, making > small adjustments for supporting the extensions and the extra-platforms we > support. I am afraid we have lost the integration with pkg-config, something > I could restore later. However, the close we track COLA's codebase the less > the maintenance costs on our end. To check it out and build, > > % gacvs co -d Grads-2.0.a8 -P Grads > % ln -s /path/to/your/supplibs/arch supplibs > % cd Grads-2.0.a8 > % configure > % make > % make check > > It passes all tests on Mac OS X 10.5.8 (Leo), x86_64 and i686. I'll try Mac > OS X 10.6 (Snow Leo) and Windows once I have robust builds on these 3 > platforms. > > 3) For testing the extensions: > > % make binstall > (put opengrads/Contents in your path) > % cd extensions > % make check > > It passes all 93 tests on Darwin and x86_64. However, it fails a single > test on i686: the creation of grads_grib files with the LATS extension (all > other formats work). I love these tests... > > Mike, > > To reproduce the problem, simply try (using lats4d.sh under > opengrads/Contents): > > % lats4d.sh -i model.nc -o junk -format grads_grib > > This will complete successfully, but when you try to read the file all > variables will be UNDEF. I tried downgrading the LATS extension to the > 2.0.a7.oga.3 version (that is, before your recent mods) but the problem > persisted. The previous 2.0.a7.oga.3 binaries work with the current LATS > extension. So it may be related to some internal grads changes, probably > related to gribmap, but I am not sure. Could you take a look at it? > > Thanks, > > Arlindo > > > > > > > > > > -- > Arlindo da Silva > da...@al... > -- Arlindo da Silva da...@al... |