Re: addr2line and the name of the executable

[Andrew Haley <aph at redhat dot com>]

Ranjit Mathew writes:
 > >  > So the property "gnu.gcj.progname" would not really be "the name by
 > >  > which the program was invoked" - isn't that (sort of) wrong?
 > >
 > >The comment only says "similar to".  It depends on what you want the
 > >information for.  If you really want the file, that's just what you
 > >need.
 > 
 > Most of the uses for argv[0] that I've seen are typically like:
 >     ...
 >     if( argc < 2)
 >     {
 >         printf( "Usage: %s <whatever>\n", argv[0]);
 >         exit( 1);
 >     }
 > 
 > In "normal" Java programs, this is replaced by
 > "Usage: java FooBar <whatever>" since this used to be the only
 > way to invoke a Java program. However, with a natively
 > compiled Java program, one cannot assume this and needs
 > access to something like argv[0]

You don't _need_ it.  It's only useful if someone does something
wicked, such as giving an executable a different name from the class
that provides the implementation.  Java does not allow this, and that
it a _very_ good thing.

 >  - this was the whole
 > point of my asking for, what became the "gnu.gcj.progname"
 > property. See this thread for the background:
 > 
 >     http://gcc.gnu.org/ml/java/2002-10/threads.html#00149
 > 
 > It is in this light that I consider the current
 > implementation to be in error.

I see.

 > If we need the actual path to the executable, this is best
 > served by something like _Jv_platform_getexecutable( ) as
 > Mohan's patch suggests.

Okay.

Andrew.