Re: [PATCH] Adding dirs for installed libraries to temp_rpath is breaking builds on darwin8

[Ralf Wildenhues]

* Albert Chin wrote on Wed, Feb 02, 2005 at 05:56:13AM CET:
> On Wed, Feb 02, 2005 at 12:57:52PM +0900, Peter O'Gorman wrote:
> > Albert Chin wrote:
> > | On Wed, Feb 02, 2005 at 08:05:55AM +0900, Peter O'Gorman wrote:
> > |
> > |>Note that Apple may change this behavior prior to an actual release
> > |>of the OS (I hope), but the patch is "harmless" anyway.
> > |
> > | Why are we supporting unreleased OS's?

This is not a good question to ask IMVHO.  If we know of certain
features that will be in upcoming OS's, we should strive to support
them.  Note that Libtool might be part of a longer development tool
chain, and starting support only after release might delay dependent
software unduly.  Example: LAM needs Libtool, provides MPI.
$simulation_code needs MPI.  Waiting for respective releases to happen
(esp. given the frequency of Libtool releases) is not fun.

OTOH, you are quite right that we should not support _speculative_
features.  (For the problem at hand, I should admit that I do not know
about speculativeness of the feature, but trust Peter's judgement.)

> > Think of it as a bugfix that also happens to make this prerelease OS
> > version work. It is better not to add the installed libs dirs to
> > DYLD_LIBRARY_PATH, doing so can cause problems even on released OSes
> > (but it is more likely to cause problems on the prerelease).
> 
> I don't like platform-specific hardcoded information in ltmain.in.
> Create a new variable in libtool.m4 (possibly similar to the existing
> hardcode_* variables) and act on that variable in ltmain.in.

ACK, thanks for mentioning this.  Creating a suitable hardcode_*
variable for this should not be hard.

Regards,
Ralf