Re: ld.so.1 broken with gcc 3.3.1 on GNU/Hurd

[Roland McGrath]

> Where can I find more info about all of this anyway?  

You'll have to be a little more specific.  You can start with the ELF spec
to understand what relocs and dynamic relocs are all about.  

> Not much of what you have written in this thread makes much sense to me
> since I have only a very vauge idea of what should happen during
> reallocation.  I am also having a hard time seeing how this is a ld bug,
> since only the gcc version was changed.  Could you shed some light on
> that?

It turns out there was no ld bug, though an objdump bug led me to believe
the contents of the file were different than they truly were and conclude
there was an ld bug.  I worked around what I thought to be an ld bug before
realizing this, so I did not actually investigate every corner of the
problem.  Off hand, I suspect the problem was ld.so containing a PC32
(i.e. non-PIC-flavor) reloc to a symbol name.