[Users] [ET Trac] [Einstein Toolkit] #100: ExternalLibraries/HDF5/HDF5.sh fails to catch a missing library file in user implementation

[Erik Schnetter]

On Mon, Dec 13, 2010 at 3:01 PM, Roland Haas
<roland.haas at physics.gatech.edu> wrote:
> Hello all,
>
>>  following:
>>
>>  make sim options=optionsfile VARIABLE=OVERRIDE
>>
>>  and thus override the setting for VARIABLE that may exist in
>>  "optionsfile".
>>
>>  We may want to change this (and I think we should); however, there is
>
> Should this really be changed? I find myself doing
> make sim options=optionsfile SILENT=NO
> occasionally when the make process fails for some odd reason (actually I
> would sometimes prefer if I could do more with this mechanism, eg. also
> OPTIMISE=NO when debugging). So if the possibility of doing
> VARIABLE=OVERRIDE on the command line is related to preferring environment
> variables over options file options then (at least for me) the current
> behaviour is the more useful one.

I think what we really want is a distinction between environment
variables and additional options that are specified on the command
line. It is unfortunate that the make command mixes them together. One
way out would be to switch from using "make" to using a script (e.g.
called "cactus") to perform the usual actions (building, configuring,
etc.).

The top-level make command (the one that the user executes) don't
actually use any real make functionality; it just parses the command
line and then performs the corresponding action. One of these actions
can be to call another make command to build Cactus; other actions are
e.g. to calls latex to build the documentation etc. This top-level
make command could be replaced by a different mechanism that would
simplify handling command line options. For example, it could also
detect misspelled (unknown) options or settings, which is basically
impossible to do from within make.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/