Re: More of that "Rune" business

On Sat, 10 Mar 2012 07:11:14 +0100
"O. Hartmann" <ohartman_at_mail.zedat.fu-berlin.de> wrote:

> On 03/10/12 06:53, Conrad J. Sabatier wrote:
> > On Fri, 9 Mar 2012 23:38:22 -0600
> > "Conrad J. Sabatier" <conrads_at_cox.net> wrote:
> > 
> >> On Fri, 09 Mar 2012 22:39:23 +0100
> >> "O. Hartmann" <ohartman_at_mail.zedat.fu-berlin.de> wrote:
> >>
> >>> On 03/09/12 21:04, Conrad J. Sabatier wrote:
> >>>> I'm getting quite a few of these "Rune"-related errors during
> >>>> port builds lately.  I've tried following the advice from the
> >>>> list, but no good, they still keep occurring.  I even tried
> >>>> backing off to my last known good buildworld/buildkernel (around
> >>>> Feb 23), and it still doesn't help.
> >>>>
> >>>> Also seeing problems relating to the new clang src.conf
> >>>> variables. Can't successfully build world and/or kernel to try
> >>>> to correct things. "make buildenv" in /usr/src hasn't helped.
> >>>> "make install" in /usr/src/share/mk hasn't helped.
> >>>>
> >>>> Is there some "magic bullet" for this that I've just somehow
> >>>> managed to overlook?  I'm totally at a loss here.  Help!
> > 
> > [failed port build output snipped]
> > 
> >>> Me, too, here.
> >>> Buidling a world with most recent sources and CLANG doesn't work,
> >>> if building with legacy/old gcc 4.2.1 the option
> >>> WITH_LIBCPLUSPLUS= YES blow off things.
> >>>
> >>> I had the "_Rune" problem quite often and came around by walking
> >>> back to a well known source base and doing a "make
> >>> installincludes".
> >>
> >> Yes, it's a very frustrating issue, because it also affects ports
> >> builds, not only source.
> >>
> >> Just on the outside chance it might help, tonight I hosed my entire
> >> src tree (which I normally update from a local copy of the CVS
> >> repository which I maintain via csup) and did an svn checkout of
> >> the src tree. Still no go.  Same exact thing.
> >>
> >> I'm completely stuck for the time being, until the root cause of
> >> this problem is corrected.  I only have about a 50/50 chance at
> >> the moment of any port builds succeeding, with numerous updates
> >> lying in wait, and zero chance of a successful world/kernel
> >> build.  I don't pretend to understand what the real issue is; all
> >> I do know is what I've been seeing lately: numerous failed ports
> >> builds, most of them referring to an unresolvable
> >> _ThreadRuneLocale, and some of them mentioning issues with yylex
> >> et al.  I'm stumped, honestly.  Been reading the lists in hopes of
> >> learning of a working solution, but nothing I've tried so far has
> >> produced any positive results.  I'm seeing the word "Rune" in my
> >> sleep lately, I swear!  :-)
> >>
> >> This is absolutely maddening!
> > 
> > Well, now, this is interesting.  Just for curiosity's sake, I tried
> > building a new kernel with the fresh source tree I just fetched from
> > the svn repository, and it succeeded!  Still can't build world,
> > though.
> > 
> > The question now is: do I dare install this new kernel and give it a
> > try?  Cant afford to do any harm to my existing installation.
> > 
> > My latest working kernel is from Feb 23.  Just how dangerous might
> > it be to try the new one?
> > 
> 
> 
> Kernel building wasn't an issue all the time. Only "world" was broken.
> And, funny, world compiles now when issuing "make buildworld". But it
> takes ages ... Even on my brand new Core i7-3930X. While compiling
> FreeBSD 10.0-CURRENT/amd64 with LLVM EXTRAS and CLANG (I do not know
> whether the whole LLVM has it made finally into the build process)
> within 30 minutes on the new lab's box, with "make buildworld", no -jX
> (usually 12 for the 6-core/12 thread box) is now heating up only one
> thread/core for about two hours.

That's kind of weird, huh?

> Well, I decided to rebuild all ports! This is easy to say and do if
> the hardware is quite potent. At my office at home, I do not dare
> doing this since the CPU is an older Intel E8500 with only two cores.
> It takes a day and more to accomplish building the ports I use to get
> rid of the "Rune"-issue. Rune seems to have a new sematics: ruin my
> day ...

"Rune" my day.  :-)  I like that!

I'm beginning to suspect it's a library issue in my case.  Gonna try
building/installing libc and friends and see if that does any good
(that is, if the build even succeeds!).

-- 
Conrad J. Sabatier
conrads_at_cox.net