Am Freitag, den 10.06.2016, 18:31 +0100 schrieb Ken Moffat:
> On Fri, Jun 10, 2016 at 05:56:16PM +0200, Thomas Trepl wrote:
> > Hi all,
> >
> > after a long time of abstinence I'll restart playing with LFS. So
> > for
> > so good as I come to chap6 until trying to install the headers.
> >
> Welcome back!
>
> Is this the last release or recent -dev ?
The book is the SVN with added cpio and openssl (patched into the book
locally)
>
> Real hardware or in a VM ? And if a VM, which one ?
Strange enough, both. I have the very same effect in a VirtualBox-VM,
The other machine is a VXL Itona with an Intel Atom N270, 2GB RAM.
>
> What host system ?
Host system on the VXL is a LFS built some times ago, kernel 3.11.
On the VM, its a LFS built in April, kernel 4.5.
>
> And, I suppose, 32-bit or x86_64 ? (in development, 32-bit gets a
> *lot* less testing.
Yes, I'd need 32-bit for the VXL installation. Tonight, I did another
run on both, my Thinkpad (64-bit) and on the VXL. Both machines run the
same scripts, host on the Thinkpad in ArchLinux.
The 64-bit runs fine, completely thru chap6-8, the VXL fails at
linux_headers. So it seems that it has something to do with 32-bit
installations.
>
> > It segfaults at "make install_headers" with
> >
> > <pre>
> > CHK include/generated/uapi/linux/version.h
> > UPD include/generated/uapi/linux/version.h
> > HOSTCC scripts/basic/fixdep
> > /bin/sh: line 1: 20146 Segmentation fault scripts/basic/fixdep
> > scripts/basic/.fixdep.d scripts/basic/fixdep 'gcc -Wp,-
> > MD,scripts/basic/.fixdep.d -Wall -Wmissing-prototypes -Wstrict-
> > prototypes -O2 -fomit-frame-pointer -std=gnu89 -o
> > scripts/basic/fixdep scripts/basic/fixdep.c ' >
> > scripts/basic/.fixdep.tmp
> > make[1]: *** [scripts/Makefile.host:91: scripts/basic/fixdep] Error
> > 139
> > make: *** [Makefile:440: scripts_basic] Error 2
> > </pre>
> >
> > I million-checked that the kernel file systems are mounted well.
> > Than i
> > tried to simply compile
> >
> > <pre>
> > #include <stdio.h>
> > int main(int argc, char *argv[]) {
> > printf("Hallo!\n");
> > return 0;
> > }
> > </pre>
> >
> > using "gcc dummy.c -o dummy" in the chroot environment. It
> > compiles/links well, but when i try to execute it, dummy segfaults
> > too.
> >
> > <pre>
> > (CH5) root:~# ./dummy
> > Segmentation fault
> > </pre>
> >
> > So it seems that the installation isn't able to build valid
> > programs in
> > chroot environment which explains why the header installation fails
> > (fixdep may also be not executable). Doing a ldd shows
> >
> > <pre>
> > (CH5) root:~# ldd dummy
> > linux-gate.so.1 (0xb7759000)
> > libc.so.6 => /tools/lib/libc.so.6 (0xb75a5000)
> > /tools/lib/ld-linux.so.2 (0xb775a000)
> > </pre>
> >
> > Where to start to search for the fault?
> >
> > Thanks,
> >
> > Thomas
> >
> Is the hardware good ? Specifically, has it been ok doing regular
> compiles recently ? I have a phenom which has good days and bad
> days, I do believe it has something screwed up in either the board
> design or its bios - but otherwise segfaults are fairly uncommon
> here, except when trying to test gcc-6.1 where all manner of things
> are breaking and a segfault can be down to miscompilation (e.g.
> openjade). But all of the gcc-6.1 problems now seem to be in BLFS,
> not on the way into chroot.
Well, i think hardware is ok, it didn't made any trouble so far and i
have the failure in a VM too.
>
> Hmm, gmp is *always* good to blame if you let it use its own
> optimised config scripts (ISTR somebody had a problem where a very
> low-end (pentium) skylake did not have the new instructions gmp
> assumed - but I think that probably failed in chapter 5).
What can i do here to ensure that gmp will not overoptimize?
On the other hand, it is not the gcc which segfaults, it is the program
created by gcc in chroot-env.
Currently I do a jhalfs build on the VXL just to see what happens.
>
> Perhaps build gdb at the end of chapter 5 ? I did that in my last
> build (--prefix=/tools using gdb-7.11) but I did not need to use it
> so no idea if it works correctly in chroot. I also built strace
> there, but gdb is probably more useful for a segfault. And gdb does
> seem to work correctly in the new system (but getting firefox to NOT
> strip its symbols seems to no-longer work).
>
> Obviously, for a script you need to run the command by hand, or
> perhaps edit the script - getting everything set up is much easier
> if you can attach gdb to a program once it is running, which I was
> able to do.
>
> Good luck!
>
> ĸen
> --
> I had to walk fifteen miles to school, barefoot in the snow. Uphill
> both ways.
--
http://lists.linuxfromscratch.org/listinfo/lfs-support
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
Do not top post on this list.
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
http://en.wikipedia.org/wiki/Posting_style
