[lfs-dev] binutils, elfutils-0.175, objdump, linux kernel-4.19.12 (SVN-20190101) Follow-up

Jean-Marc Pigeon jmp at safe.ca
Tue Jan 8 20:43:29 PST 2019


On 01/08/2019 11:16 PM, Bruce Dubbs via lfs-dev wrote:
> On 01/08/2019 08:52 PM, Ken Moffat via lfs-dev wrote:
>> On Tue, Jan 08, 2019 at 09:18:36PM -0500, Jean-Marc Pigeon via lfs-dev
>> wrote:
>>> Hello,
>>>
>>> Redone all chapter 6 with elfutils-0.173
>>> (all things equal otherwise)
>>> Kernel compilation is now OK.
>>>
>>> According:
>>> https://bugs.archlinux.org/task/61151
>>>
>>> Bug is in binutils, patch have been applied on git
>>> but there is no binutils-2.31.2 (as fare I can see).
>>>
>>> My proposal is to have SVN to back track to elfutils-0.173
>>> waiting for binutils next official release.
>>>
>>> comments?
>>>
>>
>> On the system where I most recently built LFS (but back in late
>> November) I have elfutils-0.175 and binutils-2.31.1
>>
>> I successfully built, and booted, both the 4.19.3 amd 4.20.0 kernels
>> on that machine (as well as a 4.20-rc), using - as normal - the
>> original kernel headers which were 4.19.3.  Too busy to try 5.0-rc1.
>>
>> Looking at my scripts, for elfutils I currently ensure that my
>> CFLAGS include '-g' (yeughh - my main purpose with CFLAGS is to
>> strip debug info to save space since I'm rubbish as using gdb ;)
>> and I assume that the book builds without specifying CFLAGS.  Maybe
>> that is related, maybe it is a lingering result of some previous
>> problem.
>>
>> On all my previous systems, including 8.3, I have not seen any
>> reason to update elfutils or binutils, so only that one system uses
>> 0.175.
>>
>> I don't have any objection to reverting to an older version of a
>> package (currently using an older psutils), but at the moment I
>> don't think one "doesn't build for me" (when it presumably did build
>> for whoever updated those packages) is a persuasive reason.
>
> I cannot duplicate the problem.
>
> tar -xf linux-4.19.12.tar.xz
> cd linux-4.19.12
> cp /boot/config-4.18.5 .config
> make oldconfig
> make
> make modules_install
>
> worked without problem.
>
> # CONFIG_DEBUG_INFO is not set
>
> Changing that to =y and running make caused the system to ask about
> couple of other DEBUG options (I took the defaults), but otherwise there
> were no reported problems.
>
>    -- Bruce
>
>
I agree
CONFIG_DEBUG_INFO
is the critical parameter for kernel problem

Tomorrow I shall full redo chapter 6 again with 0.175.
and double check....

What "worry" me is not the kernel compilation problem as such,
but rather an indication about objdump malfunction....
objdump is a binutil tool... (should be working).



-- 

A bientôt
===========================================================
Jean-Marc Pigeon                        E-Mail: jmp at safe.ca
SAFE Inc.                             Phone: (514) 493-4280
   Clement, 'a kiss solution' to get rid of SPAM (at last)
      Clement' Home base <"http://www.clement.safe.ca">
===========================================================

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5158 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.linuxfromscratch.org/pipermail/lfs-dev/attachments/20190108/825b172b/attachment.bin>


More information about the lfs-dev mailing list