Re: bootstrap/3236

[Jeff Deifik <jdeifik at weasel dot com>]

At 12:33 PM 6/20/01 +0100, Joseph S. Myers wrote:
>On Tue, 19 Jun 2001, Zack Weinberg wrote:
>
> > I agree this is confusing.  It's also inaccurate; the second paragraph
> > is true whether or not you are using CVS.
>
>The point of referring to CVS is that checking out from CVS will check out
>into a "gcc" directory rather than a gcc-version directory.
>
> > # % cd *objdir*
> > # % ../*srcdir*/configure [target] [options]
> >
> > to emphasize that *objdir* and *srcdir* must be entirely separate.
>
>Both relative and absolute paths for srcdir ought to work.
>
> > I would send a patch but I do not know where the master copy of this
> > file lives or how to update all of its children.
>
>gcc/doc/install.texi on mainline (which should probably be merged to the
>branch when the branch reopens), online version generated by
>update_web_docs run from cron.

This is jeff deifik (from my home domain).

I was able to successfully build gcc-3.0 by building in the
gcc-3.0 tree. I know you folks said not to do it, but someone else
at work did it successfully, and I was at wits end due to the
make bootstrap bug I reported.

The build went fine, the install went fine, and my personal test cases work 
fine.

There does seem to be a problem with the 'make -k check' part, not
being able to find a bunch of files, but I didn't report that, since I *was*
building in the source directory.

I suspect there is something wrong with the documented procedure to
have objdir != srcdir, since I was able to build when objdir == srcdir.

         thanks,
         jeff deifik
         jeff.deifik@jpl.nasa.gov - work
         jdeifik@weasel.com - home