[blfs-dev] MesaLib-10.4.3 make failed for libtool-2.4.2, solution is simple

Bruce Dubbs bruce.dubbs at gmail.com
Mon Jan 26 11:28:56 PST 2015

Fernando de Oliveira wrote:
> On 26-01-2015 15:18, Bruce Dubbs wrote:
>> Fernando de Oliveira wrote:
>>> On 26-01-2015 13:53, Bruce Dubbs wrote:
>>>> Pierre Labastie wrote:
>>>>> Le 26/01/2015 15:47, Igor Živković a écrit :
>>>>>> On 01/26/2015 03:22 PM, Fernando de Oliveira wrote:
>>>>>>> If you agree with the modifications but doesn't want to do the commit
>>>>>>> yourself, I volunteer.
>>>>>>> While waiting for your (or anybody else's) reply, I will try the
>>>>>>> "autoreconf/configure" combination (thought about it when writing
>>>>>>> this
>>>>>>> post).
>>>>>> I didn't run into this issue but then again I don't install LFS by the
>>>>>> book for a long time now. I trust you, Fernando, with making any
>>>>>> corrections necessary.
>>>>> FWIW, I just built Mesalib-10.4.3, and everything went OK : I have a
>>>>> recent version of LFS (20150112), and the installed libtool is 2.4.4.
>>>> My old -dev system has 2.4.2 and my new one 2.4.4, but I haven't build X
>>>> for the new system yet.  I can do that today if necessary and check
>>>> Mesalib instructions.  Do you need me to do that?
>>> I think both are amd64 systems, right? I don't know the architecture of
>>> Igor's and Pierre's systems. It could be interesting confirming or not
>>> that you have problem with the old system with 2.4.2, so we could be
>>> more specific in the Note. One more thing is that probably you are using
>>> GCC older than mine, there (?), mine is current 4.9.2.
>>> So, please, both tests would be much appreciated, but I believe you
>>> don't need to do it in the new machine without X yet, as they have
>>> already been done by Pierre and Igor.
>> OK, using the instructions in the book, I get:
>> make[4]: Entering directory '/tmp/mesalib/Mesa-10.4.3/src/util'
>>    CC       libmesautil_la-hash_table.lo
>> libtool: Version mismatch error.  This is libtool 2.4.4, but the
>> libtool: definition of this LT_INIT comes from libtool 2.4.2.
>> libtool: You should recreate aclocal.m4 with macros from libtool 2.4.4
>> libtool: and run autoconf again.
>> Makefile:634: recipe for target 'libmesautil_la-hash_table.lo' failed
> Yes, from memory, exactly what I got. So, it seems the instructions
> might remain as I did in r15391 and forgot to inform.
>> If I change the instructions to:
>> autoreconf -fi &&
>> ./configure  CFLAGS="-O2" ...
> Yes. the Note has that as alternative to the sed.
> I believe that just changing the default to this (with -v, perhaps)
> would make the Note unnecessary. Not sure about that though, because it
> was like that untill somebody changed to use the autgen.sh script, which
> is essentially that, but without the -f option.
> There is a note in the bad libtool generated by autogen.sh (or does it
> come bundled?) after some pages after the head recommending not to edit
> what follows, but after our experiments today, I have the impression
> that it is not really a problem generating a new one when we use
> autoreconf -fi && ./configure  CFLAGS="-O2" ...
>> It builds fine.  I didn't install though.
>> gcc-4.9.1, libtool-2.4.2, lfs SVN-20140908, x86_64.  Do you need me to
>> do a test build on a LFS-7.5 system also?
> Don't know. Seem to remember post telling that we support 2 previous
> releases. But you are a much better judge of that. Perhaps it is built
> without problem there? If you don't mind, it might be interesting.

OK, it works fine on the LFS-7.5 system with autoreconf -fi && 
./configure  CFLAGS="-O2" ...

I don't think the -v is needed.  We get enough feedback without it.

I really prefer using that to what we have now.  It's really much 
simpler.  We don't need the note at all.

   -- Bruce

More information about the blfs-dev mailing list