On Nov 16, 2011, at 3:07 PM, Dan Tenenbaum wrote: > Simo n, > > On Tue, Nov 15, 2011 at 7:26 PM, Simon Urbanek > <[email protected]> wrote: >> Dan, >> >> static hdf5-1.8.8 and netcdf-4.1.3 are now in >> http://r.research.att.com/libs/ (and also on the CRAN machine). >> > > Thanks! So far the package does not compile against that but I've > asked the package maintainer to take a look. If you are curious, the > package is ncdfFlow > (http://bioconductor.org/packages/release/bioc/html/ncdfFlow.html) and > the error I get is: > checking for nc_def_var_deflate in -lnetcdf... no > configure: error: netCDF library not found! > ERROR: configuration failed for package 'ncdfFlow' > > Seems like this might be a symbol from zlib. >
No, it's using the wrong flags to link NetCDF - it completely ignores dependencies. It has hard-coded LIBS which is simply wrong. I would suggest the author uses pkg-config to fetch the relevant LIBS, e.g. on the CRAN machine it is hagal:tmp$ pkg-config --libs netcdf -L/usr/local/lib -lnetcdf -lhdf5_hl -lhdf5 -lm -lz -lcurl so all the above is simply missing. Cheers, Simon > Dan > > >> Cheers, >> Simon >> >> On Nov 15, 2011, at 9:40 PM, Simon Urbanek wrote: >> >>> >>> On Nov 15, 2011, at 8:36 PM, Dan Tenenbaum wrote: >>> >>>> Hi all, >>>> >>>> 2011/11/15 Hervé Pagès <[email protected]>: >>>>> On 11-11-15 06:26 AM, Simon Urbanek wrote: >>>>>> >>>>>> On Nov 15, 2011, at 1:29 AM, Hervé Pagès wrote: >>>>>> >>>>>>> On 11-11-12 11:30 PM, Prof Brian Ripley wrote: >>>>>>>> >>>>>>>> On Sun, 13 Nov 2011, Riccardo Romoli wrote: >>>>>>>> >>>>>>>>> Hi, I have already installed R 2.14.0 and I have some problem to load >>>>>>>>> several packages, in particular xcms package. I installed it directly >>>>>>>>> from the GUI and when I try to load it the R I have: >>>>>>>>> >>>>>>>>>> library(xcms) >>>>>>>>> >>>>>>>>> Error in dyn.load(file, DLLpath = DLLpath, ...) : >>>>>>>>> unable to load shared object >>>>>>>>> >>>>>>>>> '/Users/riccardoromoli/Library/R/2.14/library/xcms/libs/x86_64/xcms.so': >>>>>>>>> >>>>>>>>> dlopen(/Users/riccardoromoli/Library/R/2.14/library/xcms/libs/x86_64/xcms.so, >>>>>>>>> 6): Library not loaded: /usr/local/lib/libnetcdf.7.dylib >>>>>>>>> Referenced from: >>>>>>>>> /Users/riccardoromoli/Library/R/2.14/library/xcms/libs/x86_64/xcms.so >>>>>>>>> Reason: image not found >>>>>>>>> Inoltre: Warning message: >>>>>>>>> In library(pkg, character.only = TRUE, logical.return = TRUE, lib.loc >>>>>>>>> = lib.loc) : >>>>>>>>> there is no package called ‘digest’ >>>>>>>>> Errore: package/namespace load failed for ‘xcms’ >>>>>>>>>> >>>>>>>>> >>>>>>>>> I think the error is the absence of ibnetcdf.7.dylib. Do you have any >>>>>>>>> idea how can I correct this error?? >>>>>>>> >>>>>>>> 1) Ask BioC to build the package correctly, or >>>>>>>> 2) Build from the sources yourself, preferably using a static netcdf >>>>>>>> library. >>>>>>>> >>>>>>>> People should not be distributing Mac binary packages with hardcoded >>>>>>>> paths to dependencies not in the R framework: CRAN does not. >>>>>>> >>>>>>> Not true. rggobi, Cairo and other Mac binaries on CRAN have hardcoded >>>>>>> paths to a bunch of stuff that is not in the R framework. >>>>>>> >>>>>> >>>>>> The above statement is false. >>>>>> >>>>>> All binaries that involve just libraries (like Cairo) are linked >>>>>> statically and have no dependencies (see otool -L). >>>>>> >>>>>> There are very few exceptions where it doesn't involve a library but an >>>>>> entire system: gtk, ggobi and tcltk. For those there is no other option >>>>>> since just linking them statically is not sufficient as they rely on >>>>>> other >>>>>> external files. For those we provide the binaries for download along with >>>>>> the package. >>>>> >>>>> I'm aware of that and thanks for providing those binaries. >>>>> Also thanks for confirming that Ripley's original statement was wrong. >>>>> >>>>>> >>>>>> The problem in BioC is that not only it uses unnecessary dependent >>>>>> libraries (it didn't use to when you were running it ;)), it doesn't even >>>>>> provide the needed dependencies for download with the package. That makes >>>>>> the repository useless for the user since the packages simply don't work. >>>>> >>>>> We used to use your static netcdf from here >>>>> >>>>> http://r.research.att.com/libs/ >>>>> >>>>> and that's what we should use. For whatever reason another netcdf >>>>> (non-static) ended up being installed on our build machine. We'll fix >>>>> that. >>>>> >>>> >>>> This is fixed now. At the end of our next build cycle (about 1PM >>>> tomorrow, Wednesday, Seattle time), xcms 1.30.1 will be available >>>> which is once again built against the static netcdf library (version >>>> 4.0.1) that Simon provides on his site. Other BioC packages that use >>>> netcdf (flowWorkspace and mzR) will also be fixed. >>>> >>>> Riccardo, this means that all you have to do is wait until the new >>>> package is available and then: >>>> source("http://bioconductor.org/biocLite.R";) >>>> biocLite("xcms") >>>> library(xcms) >>>> >>>> You will not need to compile anything. I just tested this on a Mac >>>> without netcdf or any compilers and it worked fine. >>>> >>>> Simon: I will be asking for your help but I want to make sure I have >>>> all my ducks in a row first. I want first to be able to build things >>>> exactly right, then I can know that I am giving you the right thing. >>>> That is taking a bit more time than I'd like. >>>> >>>> Briefly, the package in question needs netcdf (must be version 4.1.3) >>>> to be linked against hdf5 which in turn needs to link to zlib. >>>> I will contact you when I am able to create a static version of this >>>> that allows this package to compile. >>>> >>> >>> Note that in libs there is static hdf5 and netcdf 4.1.1 (which uses hdf5), >>> so the only issue is that the package needs 4.1.3 and not 4.1.1 -- I wonder >>> why since it's just a patch release ... >>> >>> Cheers, >>> Simon >>> >>> >>>> Thanks for your help. >>>> Dan >>>> >>>> >>>> >>>> >>>>> Cheers, >>>>> H. >>>>> >>>>>> >>>>>> Cheers, >>>>>> Simon >>>>>> >>>>>> >>>>>>>> In >>>>>>>> particular, /usr/local/lib need administrative privileges to populate. >>>>>>> >>>>>>> Note that installing the external stuff needed by rggobi, Cairo etc >>>>>>> also requires administrative privileges. >>>>>>> >>>>>>> Cheers, >>>>>>> H. >>>>>>> >>>>>>>> >>>>>>>>> >>>>>>>>> Best >>>>>>>>> >>>>>>>>> Riccardo >>>>>>>>> _______________________________________________ >>>>>>>>> R-SIG-Mac mailing list >>>>>>>>> [email protected] >>>>>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> _______________________________________________ >>>>>>>> R-SIG-Mac mailing list >>>>>>>> [email protected] >>>>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac >>>>>>> >>>>>>> >>>>>>> -- >>>>>>> Hervé Pagès >>>>>>> >>>>>>> Program in Computational Biology >>>>>>> Division of Public Health Sciences >>>>>>> Fred Hutchinson Cancer Research Center >>>>>>> 1100 Fairview Ave. N, M1-B514 >>>>>>> P.O. Box 19024 >>>>>>> Seattle, WA 98109-1024 >>>>>>> >>>>>>> E-mail: [email protected] >>>>>>> Phone: (206) 667-5791 >>>>>>> Fax: (206) 667-1319 >>>>>>> >>>>>>> _______________________________________________ >>>>>>> R-SIG-Mac mailing list >>>>>>> [email protected] >>>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac >>>>>>> >>>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> Hervé Pagès >>>>> >>>>> Program in Computational Biology >>>>> Division of Public Health Sciences >>>>> Fred Hutchinson Cancer Research Center >>>>> 1100 Fairview Ave. N, M1-B514 >>>>> P.O. Box 19024 >>>>> Seattle, WA 98109-1024 >>>>> >>>>> E-mail: [email protected] >>>>> Phone: (206) 667-5791 >>>>> Fax: (206) 667-1319 >>>>> >>>>> _______________________________________________ >>>>> R-SIG-Mac mailing list >>>>> [email protected] >>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac >>>>> >>>> >>>> >>> >> >> > > _______________________________________________ R-SIG-Mac mailing list [email protected] https://stat.ethz.ch/mailman/listinfo/r-sig-mac
