[uClibc] Re: RE: [uClibc-cvs] svn commit:trunk/uClibc/ldso/ldso:mipsel

[Joakim Tjernlund]

> 
> On Thu Mar 17, 2005 at 06:20:03PM +0100, Joakim Tjernlund wrote:
> > > On Thu Mar 17, 2005 at 05:31:37PM +0100, Joakim Tjernlund wrote:
> > > > > On Thu Mar 17, 2005 at 05:22:26PM +0100, Joakim Tjernlund wrote:
> > > > > > Erik, can you look at this also? Should I commit and try fixing from there? It would
> > > > > > be much simpler to work out of SVN.
> > > > > 
> > > > > In the recent flurry of emails, I must admit I've lost the
> > > > > sequence of what additional patches must be applied....
> > > > > 
> > > > >  -Erik
> > > > 
> > > > Yes, that my point. So I commit what I have and we go from there?
> > > 
> > > I suppose so...  As long as we immediately proceed to get things
> > > fixed up and booting again.
> > 
> > OK, now I have commited everything. The only arch that is known
> > to be broken is MIPS. I added yet some new assembler from glibc for MIPS, I think
> > thats whats broken. Currently there are 3 different versions of startup code
> > for MIPS, the extra versions will be removed once MIPS is working
> > again.
> 
> I'm afraid mips does not emit so much as a single line of output,
> even with SUPPORT_LD_DEBUG and SUPPORT_LD_DEBUG_EARLY enabled.
> 
> Prior to your latest commit, mips was running as expected, though
> of course one could not execute ld-uClibc-0.9.27.so.

That matches what Peter has reported as well, but I can't quite figure out what
went wrong. I think it has something to do with arch startup code in dl-startup.h.

The first asm segment(between the first #if 0 .. #endif) is almost the orginal code.
the second (or possibly the third) one is what I would prefer.

hmm, maybe the changes to GET_ARGV dosen't work for some reason. I think
the simplest way is to fir up gdb on ldso and follow the first few instructions to
find the problem.