libgcc_s is not linked in applications. Any suggestion?

William Zhou armageddonmagic at 163.com
Wed Feb 22 13:59:16 PST 2006


Hello,

There is no obvious benefit for doing this. I am a LFS user and I like 
it because it tells me how to build a Linux. I personally do not 
emphasis the time I spend on
building the system as it already took me days and nights. What I am 
trying to do is to separate the building of cross compiler and the 
temporary/final system. Clearly,
the book is divided into three parts. Make the Cross-Compile Tools, 
Building the Basic Tools and building the LFS System. It would be nice 
if we did this according
to this logic.

As you stated, there is no glibc in /tools. There is no glibc in 
/cross-tools before we built it. So we can build one in /tools. IMO,  
out temporary system build should not be
mixed with cross compiler. I understand that it took you hours to build 
a system to test something, but it is not what LFS for, LFS is for 
teaching people how to build Linux.

Back to my problem, the only issue I have currently is that programs 
compiled by the cross-compiler cannot locate libgcc_s.so correctly.  If 
I copy/move the libgcc_s.so* to
directory /ctools_i686-i586.2/lib, everything works as expect. From the 
log we can see that the i586-pc-linux-gnu-gcc did found the files in 
directory /ctools_i686-i586.2/i586-pc-linux-gnu/lib.
Then why the program have to search /ctools_i686-i586/lib instead of 
/ctools_i686-i586.2/i586-pc-linux-gnu/lib? I have zero knowledge in 
C/C++ and I cannot do much.

William


Ken Moffat Wrote:
> Without other changes, this cannot work - there is no glibc in /tools. 
> Possibly, you can interpose another build of glibc after item 6, but 
> what tangible benefit are you hoping to achieve ?
>
>  On multilib, you have to build glibc twice each time - making people 
> do that another time had better have some meaningful benefit, 
> particularly when we remember that all of /cross-tools and /tools is 
> temporary stuff that we can't even run the testsuites for.  Today I 
> wanted to test a different glibc snapshot on ppc64, towards fixing an 
> apparent bug in ld which I suspect prevents firefox from compiling - 
> five and a half hours of building to be able to run the binutils 
> testsuite, which wasn't any better.  Add in a few more compiles of 
> glibc, and perhaps gcc, before I can run the testsuite and I will not 
> be happy 8-|
>
> Ken






More information about the cross-lfs mailing list