On Thu, Sep 24, 2015 at 06:50:38PM -0400, Wayne Sallee wrote:
> Anyone able to build atk?
> 
Yes, thanks.  And whoever put that version into whichever version of
the book you are using (in svn, atk is 2.16.0, with corresponding
changes throughout the gtk+-2 and gtk+-3 stacks) was also successful
at that time.  And I'm sure I've built atk-2.14.0 in the past.

> I'm getting:
> make[3]: Entering directory '/usr/src/atk/atk-2.14.0/atk'
>   CC       atkaction.lo
>   CC       atkcomponent.lo
>   CC       atkdocument.lo
>   CC       atkeditabletext.lo
>   CC       atkgobjectaccessible.lo
>   CC       atkhyperlink.lo
>   CC       atkhyperlinkimpl.lo
>   CC       atkhypertext.lo
>   CC       atkimage.lo
>   CC       atknoopobject.lo
>   CC       atknoopobjectfactory.lo
>   CC       atkobject.lo
> atkobject.c: In function 'atk_object_class_init':
> atkobject.c:406:57: internal compiler error: Illegal instruction
>                                                          G_MAXDOUBLE,
>                                                          ^
> 0x87815f crash_signal
>     ../../gcc-4.9.2/gcc/toplev.c:337
> 0x7fa16652651f ???
>     
> /usr/src/glibc/glibc-2.21/signal/../sysdeps/unix/sysv/linux/x86_64/sigaction.c:0
> 
> 
> Wayne Sallee

An ICE (internal compiler error) means something *apparently* went
wrong in your compiler.  On one of my machines (an AMD phenom - they
are notorious for this), building with -j4 often provokes this sort
of error, particularly when building a new kernel.

Usually, asking "the big fella" to run "echo 3 >
/proc/sys/vm/drop_caches" solves it, but other times I have to use
-j3 (on a machine with 4 CPUs).  But last night I again had that
problem, when rebuilding sqlite3 for firefox-41, and it seemed to be
happening in code I had previously not built and dropping the caches
did not help (sqlite3 is already make -j1).  In that case (an LFS-svn
from late June), dropping my forced CFLAGS from -O3 to -O2 solved it.

More generally, an ICE can be caused by overheating (including dust
on the fan), bad power, bad RAM.  If dropping the caches, and/or
dropping back to make -j1, does not solve it you might have a fun
time fixing it.  Memtest86+ might be useful :-(

To be fair, some ICEs really do indicate a compiler error.  With
such an old compiler (this is LFS, we used to be bleeding edge, and
for things like gcc we are still near it) that seems unlikely, at
least for packages in BLFS, unless you built it with exotic CFLAGS.

ĸen
-- 
Il Porcupino Nil Sodomy Est! (if you will excuse my latatian)
  aka "The hedgehog song"
-- 
http://lists.linuxfromscratch.org/listinfo/blfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to