which gets mapped to a specific address for Alpha binaries by the
Fine, then _tell_ me what the specific address is supposed to be,
if there even really is one. I know there's a specific address in
SVR4/i386, but as yet you've given me no reason to think the address
is wired down on the Alpha. Just saying it's wired down doesn't do
anybody a bit of good if you don't state what the address is supposed
Do you have a copy of DEC UNIX with a compiler?
Write a program:
And then examine the memory mappings for it.
I can't really *prove* that the thing *never* maps it to a different
location, but I've put loops like that into much more complex programs
and not had it move around.
not the crt0.o (that's why you can dlopen from static executable on
an Alpha; same for SVR4 and Solaris, for that matter).
As I've told you at least four times in the mailing lists, that is
misinformation. I know from first-hand experience (and from man
pages) that you can't use dlopen in static executables under SVR4.
And I and others have quoted every variety of Solaris man page to you,
which specifically say you can't do it. I wish you'd quit spreading
it around that you can, because it's utterly false.
There isn't a static "libdl".
You're right. I screwed up and assumed that the libdl was statically
Hmmm... you *can* use mmap and libelf to get the offsets, right? So
it is possible to support loadable code modules in a static binary,
it's just an implementation decision that that's not how dlopen
itself is implemented.