Dan- for future reference, could you also
submit requests like this to the netcdf support
so we can track them in our database.
In any case, the problem is as you say with bison and flex
naming; I just fixed it and it should be available tomorrow
in tonights snapshop. Let me know if there a continuing problem.
=Dennis Heimbigner
Unidata
Dan Kokron wrote:
There is a namespace collision between mvapich2 and NetCDF4 when NetCDF
(4.1) is compiled with parallel HDF5 (1.8.4). Seems both mvapich2 and
NetCDF use flex/bison to generate some code and the generated routine
names collide.
E.g.
/nobackup/***/mvapich2-1.4/lib/libmpich.a(parser.o): In function
`parser_setup_string':
/home1/***/***/mvapich2-1.4/src/mpid/ch3/channels/mrail/src/plpa/parser.l:77:
multiple definition of `yyin'
ncgentab.o:/home1/***/Baselibs/GMAO-Baselibs-3_1_5/src/netcdf-4.1/ncgen3/ncgen.y:571:
first defined here
/nobackup/***/mvapich2-1.4/lib/libmpich.a(parser.o): In function
`parser_setup_string':
/home1/***/***/mvapich2-1.4/src/mpid/ch3/channels/mrail/src/plpa/parser.l:77:
multiple definition of `yyout'
ncgentab.o:/home1/dkokron/Baselibs/GMAO-Baselibs-3_1_5/src/netcdf-4.1/ncgen3/ncgen.y:570:
first defined here
See
http://mail.cse.ohio-state.edu/pipermail/mvapich-commit/2009-October/001080.html
for more information.
I've searched the netcdfgroup archives for this issue without finding
anything. Has anyone else seen this?
