Bk6.6Ch6.9 glibc `__stack_chk_guard' work around.

x2164 at lycos.com x2164 at lycos.com
Mon Mar 29 22:49:23 PDT 2010

----- Original Message -----
From: Bruce Dubbs <bruce.dubbs at gmail.com>
To: LFS Support List <lfs-support at linuxfromscratch.org>
Sent: Tue, 30 Mar 2010 00:25:37 -0400 (EDT)
Subject: Re: Bk6.6Ch6.9 glibc  `__stack_chk_guard' work around.

x2164 at lycos.com wrote:
>      Hi,
>      This is the work around i'm using.
>      At this point i resumed compiling at Bk.6.6_Ch.6.17, Sed,
>      and have so far completed through Ch.6.59, Vim, without
>      errors.

>This indicates to me that you made an error in Chapter 5.  The error was 
>probably in either gcc or glibc.
>When we prepare for a LFS release, we have automated scripts that 
>extract the instructions from the book and build the entire system. 
>Your work around is not necessary if all the instructions are followed 
>Your comment is that you "pretty much" followed the book.  Either you 
>made purposeful changes or missed an instruction.   If you made *any* 
>intentional deviations, please elaborate them.
>In either case, I wouldn't trust your system.  I recommend starting 
>over.  Of course, it's your distro, so you make the rules for your system.

>   -- Bruce


     I threw in the "pretty much" this time as a freebie given the
     controversy it caused last time.

     Started Bk.6.5 just as Bk.6.6 was released.  Had gotten to Bk.6.5Ch.6.9
     and had to stop because of the `__stack_chk_guard' error.  Thought i
     could have made an error so i 'rm -rf /mnt/lfs/*' and re-started using
     then newly released Bk.6.6 being especially careful to follow the 
     instructions to the letter.  Of course as you know from my previous post
     i encountered the `__stack_chk_guard' error again in Bk.6.6Ch.6.9.

     I think your right about the error being in gcc.  The libssp and gcc
     were both installed at the same time, both have same time stamp.  Why
     gcc couldn't find it, well, i don't know.  The -fstack-portector option
     was so suppose to pull in libssp, at least that was my understanding,
     so that -lssp wasn't needed for final linking line.

     Since i don't have the scripts to automate the complete build but instead
     have to do it manually from the beginning, which is kinda time consuming,
     re-starting to find the source of the error isn't feasible for me.

     Given that this error, `__stack_chk_guard', only cropped up for the
     production of the 'nscd' executable with the 2nd pass gcc, and, not
     at all with the Ch.6.16 gcc, i feel confident enough, especially with
     both resulting glibcs' passing their test suites, to use the system
     as it is now.

     And if it falls apart and bites me on the a** then i'll have afforded
     you one of the supreme pleasures in life.  The right to tell me
     you told me so. :-)

     I don't mean this to be critical, but, since your using automating 
     scripts to generate your version of LFS it comes across as being
     kinda of hypocritical to say to we who do this manually that we
     must suffer from the beginning to obtain our pure and worthy distro
     of LFS.  I'm not even going to try and point out that the scripts
     are adding their own unknown variability to difference between manual
     and automated production.

     Time to sleep.  Won't be able to check for responses till after 9 EDT.


     pete x2164

More information about the lfs-support mailing list