Re: buildworld failure r223619 to 225128

On Thu, Sep 1, 2011 at 6:49 PM, Beach Geek <labeachgeek_at_gmail.com> wrote:

> On 8/27/11, Matt Thyer <matt.thyer_at_gmail.com> wrote:
> > Have you tried moving aside your existing /usr/src and checking it out
> > afresh?
> > On Aug 28, 2011 12:47 AM, "Beach Geek" <labeachgeek_at_gmail.com> wrote:
> >> On 8/25/11, Dimitry Andric <dim_at_freebsd.org> wrote:
> >>> On 2011-08-25 17:12, Beach Geek wrote:
> >>>> make buildworld failed trying to upgrade from r223619 to r225128.
> >>>> (Note: Updating other boxes from r224774 to r225119 went flawless)
> >>>>
> >>>> On failing laptop (Toshibs Sat C655D)
> >>> ....
> >>>> /usr/include/c++/4.2/bits/stringfwd.h:56: internal compiler error:
> >>>> Segmentation fault: 11
> >>>> Please submit full report,
> >>>
> >>> That is most likely a hardware problem. Please run a full memtest,
> >>> and/or any other hardware diagnostics you can find.
> >>>
> >>> It could also be running out of memory, but that is less likely, and
> you
> >>> usually get another signal then. But who knows what might happen if you
> >>> choke a compiler. :)
> >>>
> >>>
> >>>> I do "rm -r /usr/obj/*" and make clean (in /usr/src) before doing
> >>>> buildworld on all boxes.
> >>>> I also tried compiling new GENERIC kernel then doing buildworld. It
> >>>> failed with same message.
> >>>
> >>> It dies on exactly the same file?
> >>>
> >>>
> >>>> Reverted to old/original kernel and tried "make depend" in /usr/src.
> >>>
> >>> You can't do that, you must run buildworld.
> >>>
> >>>
> >>>> It failed with... (by hand again)
> >>>>
> >>>> ===> lib/clang/libllvmarmasmparser (depend)
> >>>> tblgen -l
> >>>>
> >
> /usr/src/lib/clang/libllvmarmasmparser/../../../contrib/llvm/lib/Target/ARM
> >>>> -I
> /usr/src/lib/clang/libllvmarmasmparser/../../../contrib/llvm/include
> >>>> -I
> > /usr/src/lib/clang/libllvmarmasmparser/../../../contrib/llvm/lib/Target
> >>>> -gen-asm-matcher -o ARMGenAsm Matcher.inc.h
> >>>>
> >
> /usr/src/lib/clang/libllvmarmasmparser/../../../contrib/llvm/lib/Target/ARM/ARM.td
> >>>>
> >>>> tblgen: Record 'CCR', field 'MemberList' does not have a list
> > initializer!
> >>>> *** Error code 1
> >>>> Stop in /usr/src/lib/clang/libllvmarmasmparser.
> >>>
> >>> Yes, this is expected. When you do not use the buildworld target, the
> >>> tblgen used above will be run from /usr/bin, which is too old. This is
> >>> why buildworld first builds an up-to-date tblgen under /usr/obj, and
> >>> uses that to generate the needed files.
> >>>
> >>
> >> This laptop also runs MS Win 7/64 and FreeBSD 9 amd. The FBSD amd
> upgraded
> > ok.
> >>
> >> The buildworld always fails in same place, with same message (5 tries).
> >> I'm running diags on it right now just to make sure the hardware's good.
> >>
> >> The reason I tried "make depend" was because of a reference to r221543
> >> that said it required "make depend" before buildworld. (a shot in the
> >> dark before posting to mail list).
> >>
> >> I will post if I find any hardware problems.
> >>
> >> Thanks,
> >> Beach Geek
> >>
> >> PS. Option on updating to a version inbetween, then to latest???
> >> _______________________________________________
> >> freebsd-current_at_freebsd.org mailing list
> >> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> >> To unsubscribe, send any mail to "
> freebsd-current-unsubscribe_at_freebsd.org"
> >
>
> Just an update:
> 1) Hardware diagnostics showed no problems.
> 2) Backed up and deleted contents of /usr/src and copied it from a box
> that upgraded ok.  Still fails same place.
> 3) Deleted contents of /usr/src again, loaded seed, and "svn up
> /usr/src". Failed and rebooted.  I'll be running buildworld again and
> try to catch msg before reboot.
>
> Thanks,



> Beach Geek
>

Just to close this thread out...
After quite a few more svn upgrades and buildworlds, and try the suggestions
in this thread and sent directly to me....  I gave up.  :(

Final solution, boot from memstick image, use DESTDIR= with
buildworld/buildkernel/etc, upgrade to 9-beta3 went fine.  Then "svn up
/usr/src to lastest, and buildworld/buildkernel/etc again... life is good.
;)

Would have been nice to know why buildworld was failing before, but using
the memstick got me current with all my data intacted.

Thanks for all the suggestions and help.
Beach Geek

PS. A side note, we got a native driver working for the RTL8188CE wireless
card, but it's still really buggy, but it's a start.