Re: win95/cygwin32 bootstrap failure

[Mumit Khan <khan at xraylith dot wisc dot edu>]

On Wed, 29 Jul 1998 N8TM@aol.com wrote:

> I've been looking into the inability to make the stage1 compiler under
> win95/cygwin32.  If I eliminate the redefinition of GET_ENVIRONMENT in xm-
> cygwin32.h, which fails because cygwin32_posix_path_list_p is not defined in
> win95, the next problem is that collect2 can't find ld(.exe).  I've tried all
> sorts of things, including editing Makefile to have the full path specified
> for LD.
> 

As always, it would be useful to have a bit more info on how/what fails.
Are you saying that the current definition of GET_ENVIRONMENT fails? Also,
what you're seeing after eliminating GET_ENVIRONMENT doesn't make sense
unless you're not properly cleaning before rebuilding since that's the
only place where cygwin32_posix_path_list_p is used.

I never build on Win95, so it would help if you would supply the details 
and hopefully we can come up with a fix. I do know that it builds under NT 
4.0 SP3.

Now, the issue of collect2 is a whole different one. The original problem
was that collect2 was also called "ld", and that caused problems. Now that
it's called collect2, that particular problem is solved. Also solved is a
bug in freopen in cygwin that caused the error messages to disappear.
Could you try running with -Wl,-debug -Wl,-verbose and see what is really
happening? (Note: my local patches disable collect2 for now on *win32; when
we revisit the issue of dwarf2 unwind based EH, I may reenable it. The
only other advantage of collect2 is -frepo, which is totally unnecessary
on ELF and PE-COFF IMO).

Regards,
Mumit