Wayne,
Yes I think you have the same understanding of the status quo as I do.
But unless I misunderstand you, I do not think your solution addresses
the weakness that I pointed out.
In a simple scenario, say you wanted to have the binaries in one place,
say on a network drive, and you wanted to install the data files, such
as libraries, etc elsewhere, the fact that these are tied to the same
root tree is a restriction that is too harsh in my opinion. It is
common in linux/unix to have two separate trees one for app binaries and
one for app data.
I simply following a time honored trend. To do it, there needs to be a
separate anchor path for the data part of the tree, not tied to
CMAKE_INSTALL_PREFIX.
CMAKE_INSTALL_BINARIES_PREFIX and CMAKE_INSTALL_DATA_PREFIX or similar.
Why is this time honored?
Only an administrator might have authority to update the program files,
whereas each user might be given authority to control his own data.
This kind of thinking comes not from me. I believe some of it is
discussed in the LSB document.
It is simply a matter of adding and decoupling CMAKE variables. They
can default to the same however. Then we need to change where the data
files are found. This is either autogenerating a hard coded path at
installation time, or by testing an environment variable.
Another scenario is when somebody, say a developer, wants multiple
binaries, but only a single copy of the libraries and application data.
I misunderstood the problem. I was looking at the problem from
installing both the binaries and the data in non-standard location for
development purposes as opposed to using the standard data with
development binaries. As of now, if the standard library paths are not
found, you get a bunch of library not found messages.
This is not a lot of work, and I think it would add value to a user's
experience.
I think the environment variable is workable as long as it supports
multiple paths. This would give the user the option to define the
library path search order as well.
