Re: Patch: java/lang/natRuntime.cc

[Andrew Haley <aph at redhat dot com>]

Ranjit Mathew writes:
 > Andrew Haley wrote:
 > > Michael Koch writes:
 > >  > > > +      char* libraryPath = ::getenv ("LD_LIBRARY_PATH");
 > >  > >
 > >  > > Sorry, but I don't think this is portable. For example,
 > >  > > on Windows, PATH is used to search for DLLs. My memory
 > >  > > is a bit hazy now, but on AIX it was something like
 > >  > > LD_LOAD_LIBRARY_PATH.
 > >  > 
 > >  > Well, getenv() is used in one other place in this file too. Thats why 
 > >  > I decided to use this. I can write a patch included the AIX variable 
 > >  > later today
 > > 
 > > If I recall correctly, Darwin uses something different again.  
 > 
 > It seems to use "DYLD_LIBRARY_PATH":
 > 
 > http://developer.apple.com/documentation/Darwin/Reference/ManPages/html/dyld.1.html
 > 
 > 
 > > Well, I have to ask what this is for.  Any gcj application that uses
 > > this will be fragile and have reduced portability.  The shared library
 > > loader searches its path using its own algorithms.  Why does a gcj
 > > program need to know this stuff as well?
 > 
 > For one, this is one of those mandatory System properties that a
 > JRE should be able to return (see JavaDoc for
 > System.getProperties()). Maybe it is used by Sun's implementation
 > of System.loadLibrary( ) and for some reason they thought it a
 > useful property to expose. (For example, if a Java application is
 > not able to load its JNI library, it can crib usefully asking the
 > user to put the library in one of the folders in this path...)

Sigh.  This one seems quite bizarre, but I guess we're stuck with it.

Andrew.