Re: [Sbcl-devel] compile error with current CVS

[William H. N. <wil...@ai...>]

On Sat, Dec 22, 2001 at 03:47:59PM +0000, Christophe Rhodes wrote:
> On Sat, Dec 22, 2001 at 02:33:39PM +0300, Alexey Dejneka wrote:
> > Hello,
> > 
> > Nathan Froyd <fr...@ro...> writes:
> > 
> > > While compiling src/compiler/generic/objdef.lisp:
> > > 
> > > debugger invoked on condition of type UNBOUND-VARIABLE:
> > >   The variable SB-KERNEL::STRUCTURE-STREAM is unbound.
> > 
> > I saw this message compiling 0.pre7.92 with 0.pre7.78, but 0.pre7.73
> > makes no problem. Try to build with earlier version.
 
I never saw the bug reported by NF in my progress through those
versions (mostly compiling sbcl-0.pre7.x with sbcl-0.pre7.x-epsilon
where epsilon is small) but there were some pretty shaky versions in
there which were just good enough to compile themselves but could have
had problems compiling future versions, and I may've mistakenly
checked something in which wasn't even good enough to compile itself.

I just rebuilt sbcl-0.pre7.97 (current CVS version) with itself
successfully, so I hope the bug is gone now.

> I've had occasional building troubles with .pre versions, too; some
> are solved by having fresh CVS checkouts (no, I don't understand
> either), 
 
Some of the .pre7 versions have been more messed up than the usual
main branch version. I try to avoid checking in such problems on the
main branch, but sometimes I screw up, and I've stumbled into more
deep problems in .pre7 than I usually do.

> but for persistent cases I find that building with CMUCL is
> fairly reliable.

That seems to be a good approach. When I have bootstrap problems I
usually start over from sbcl-0.6.13, since CMU CL doesn't run on my
main (OpenBSD) machine, but the principle is the same.

-- 
William Harold Newman <wil...@ai...>
HAPPY HACKING TO ALL, AND I'LL BE BACK NEXT YEAR!
  -- http://www.yetanother.org/damian/diary_December_2001.html#day_24
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C