[BioC] Building Rgraphviz w/ Graphviz 2.14

Emil Lundberg Emil.Lundberg at bmc.uu.se
Fri Sep 21 12:52:32 CEST 2007


Hi all,

I have trouble installing/compiling Rgraphviz 1.14.1 for BioC 2.0.  
I'm on Mac OS X 10.4.1 server (Intel). This thread started on the  
MacPorts mailing list (providers of the Graphviz package for OS X),  
then took a side tour to the Graphviz developers, but I now believe  
it belongs here.

Response from macports-users: http://lists.macosforge.org/pipermail/ 
macports-users/2007-September/005630.html
Response from graphviz-devel: https://mailman.research.att.com/ 
pipermail/graphviz-devel/2007/000450.html

The root of the problem is that graphviz >=2.14 is missing some  
header files as compared to older versions. I assume this is due to  
the cleanup of the code rather than missing functionality, but in any  
case, a revision of the Rgraphviz package is probably in order.

Until an Rgraphviz update hits the trunk, do you have any tips on how  
I can get the current Rgraphviz to compile manually? Or dare I try  
the OS X binary? Se below for compilation specifics.

TIA,

Emil Lundberg
MicroArray Datawarehouse Resource (MADR)
Linnaeus Centre for Bioinformatics (LCB)
Biomedical Centre (BMC)
Uppsala University


**********
(manually in the Rgraphviz directory, or through bioclite("Rgraphviz"))
% R CMD INSTALL .
* Installing to library '/opt/local/lib/R/library'
* Installing *source* package 'Rgraphviz' ...
configure: No --with-graphviz option was specified. Trying to find  
graphviz using other methods.
checking for pkg-config... /opt/local/bin/pkg-config
configure: Found graphviz 2.14.1
configure: We have inferred the following flags for graphviz
configure: ..PKG_CPPFLAGS=-I/opt/local/include/graphviz
configure: ..PKG_LIBS=-L/opt/local/lib -lgvc
configure: creating ./config.status
config.status: creating src/Makevars
** libs
/opt/local/bin/gcc-mp-4.2 -std=gnu99 -I/opt/local/lib/R/include -I/ 
opt/local/lib/R/include -I/opt/local/include/graphviz    - 
DGRAPHVIZ_MAJOR=2 -DGRAPHVIZ_MINOR=14 -I/opt/local/include  - 
D__NO_MATH_INLINES  -fPIC  -O2 -c LL_funcs.c -o LL_funcs.o
In file included from /opt/local/include/graphviz/gvc.h:20,
                  from common.h:30,
                  from LL_funcs.c:1:
/opt/local/include/graphviz/types.h:27:1: warning: "TRUE" redefined
In file included from common.h:5,
                  from LL_funcs.c:1:
/opt/local/lib/R/include/Rdefines.h:148:1: warning: this is the  
location of the previous definition
In file included from LL_funcs.c:1:
common.h:41:20: error: gvcint.h: No such file or directory
common.h:42:21: error: globals.h: No such file or directory
make: *** [LL_funcs.o] Error 1
chmod: /opt/local/lib/R/library/Rgraphviz/libs/*: No such file or  
directory
ERROR: compilation failed for package 'Rgraphviz'
** Removing '/opt/local/lib/R/library/Rgraphviz'

I've already tried removing the gvcint.h and globals.h includes from  
common.h, but I then get:

[...]
graphvizVersion.c: In function 'Rgraphviz_graphvizVersion':
graphvizVersion.c:24: error: dereferencing pointer to incomplete type


**********

Thanks guys for adressing this. My graphviz (2.14-1 from MacPorts) is  
missing the following headers compared to Ryan's 2.12:

aghdr.h
agxbuf.h
arith.h		(back as of 2.14-3)
ast_common.h
dthdr.h
exgram.h
exlib.h
expr.h
globals.h	(requested by Rgraphviz)
grammar.h
gvcint.h	(requested by Rgraphviz)
libgraph.h
logic.h
malloc.h
parser.h
pathplan.h
pathutil.h
solvers.h
tri.h
triefa.h
vis.h
vispath.h
vmstub.h

To clarify matters, the Rgraphviz is an external R package that  
depends on the Bioconductor package, so the R developers might not  
know about it. R builds fine using the current MacPorts version  
(don't know about the devel), and so does Bioconductor, but neither  
depend on graphviz as such.

The maintainers of Rgraphviz have already made a good effort in that  
package's header files to accomodate different versions of graphviz  
(up to 2.10 at least), so this is hopefully what they need to know to  
update the Rgraphviz package. I'm forwarding this to the bioconductor  
mailing list.


best,

/Emil



More information about the Bioconductor mailing list