HDF5 dilemma

[Mark Moll]

HDF5 1.6.10 and HDF5 1.8.4 are currently in prerelease. HDF5 1.6.10 is  
the last release of the 1.6.x series. HDF5 1.8.x can also be compiled  
in 1.6.x compatibility mode, but this shouldn’t be done by default. I  
think it’s hard to make a case for the hdf5_select approach you  
suggest. First, there’s only 2 or 3 ports that need hdf5 1.6 (I  
think). Second, with gcc and python there really are many versions  
simultaneously in use. The cost of switching versions is significant  
with gcc and python and it makes sense to support multiple versions.

Since I am the maintainer of the hdf5-18 port I might be somewhat  
biased, but the least bad solution might be to have the hdf5 port  
install its files in ${prefix}/lib/hdf5-16/.

On Nov 6, 2009, at 5:34 PM, Darren Weber wrote:

>
> I've got a hdf5 dilemma ;-)
>
> $ port installed hdf5*
> The following ports are currently installed:
> hdf5 @1.6.9_0+threadsafe (active)
> hdf5-18 @1.8.3_0
> hdf5-18 @1.8.3_1
>
> $ sudo port activate hdf5-18 @1.8.3_1
> --->  Activating hdf5-18 @1.8.3_1
> Error: port activate failed: Image error: /opt/local/bin/gif2h5 is  
> being used by the active hdf5 port.  Please deactivate this port  
> first, or use 'port -f activate hdf5-18' to force the activation.
>
> The hdf5 and hdf5-18 ports are behaving like separate ports, up to  
> the point of activation conflicts.  There are two maintainers for  
> these ports (in the CC list of this email); can we get together on  
> this and work out the activation conflict?
>
> Is it possible to have multiple version specific libs/bins  
> installed?  Is it as simple as providing some version specific file- 
> name mangles (with symlinks and maybe a hdf5_select utility like the  
> gcc_select or python_select utility)?
>
> A quick search on the user email list brings up a number of ports  
> that depend on hdf5 with dependency build issues.
>
> What is the current status of play on hdf5 and what is the  
> recommended version to have installed?
>
> Take care,
> Darren
>

-- 
Mark



-- 
Mark



-- 
Mark