[PATCH] Darwin - verbose linker messages influence configure results.

[Peter O'Gorman]

Hi,

Jeremy reported this ages and ages ago, and I promptly forgot about it 
(sorry). David reported it again yesterday.

Darwin's linker can be quite verbose, if it sees an -L path that does 
not exist it prints a warning on stderr. Jeremy needs to configure with 
-L paths that don't exist at configure time (because they will exist 
later), and would like to be able to debug the result, so detecting 
-force_load properly would be helpful.

Davids problem was about failing to detect -single_module, same issue.

I only have Mac OS X 10.6 systems, and I am much too lazy to check what 
older linkers did when they saw flags they didn't recognize, so this 
patch greps stderr for "flag" and assumes the linker didn't recognize it 
if is there. I think all Darwin ld's exit with an error for flags they 
don't recognize, so even that is probably not necessary, but I am not 
100% sure of that.

Jeremy, I set RC_TRACE_ARCHIVES, RC_TRACE_DYLIBS, 
RC_TRACE_DYLIB_SEARCHING to check that none of those env vars caused ld 
to print its arguments on stderr, and they didn't. Are there any env 
vars set when building the OS that would cause that to happen? If so I 
will change this patch to just assume that ld exits with an error code 
for unrecognized arguments.

Patch Ok?

Peter

0001-Don-t-let-verbose-linker-messages-influence-test-res.patch
Description: Text Data