On Thu, Mar 10, 2005 at 03:56:27PM -0800, Per Bothner wrote:
Daniel Jacobowitz wrote:
So does it still help so much?
I think so, but I'll have to re-compile the library.
I did try a simple hello world:
-rw-rw-r-- 1 bothner bothner 56053 Mar 10 15:33 Foo1.s # before
-rw-rw-r-- 1 bothner bothner 12607 Mar 10 15:39 Foo2.s # after
Is the resulting debug information useful?
Why not? Of course it needs to be tested, but the mechanism works
for C++,
and has been used by gcc/gdb for many years.
I know the principle is sound, although it's less "obvious" than you're
making it sound - for instance, the user is SOL if a Java library lacks
debugging info. I was asking about the Java-specific bits, i.e.
whether the resulting set of output classes is appropriate. I'll just
take your word for it; I barely even speak Java.