__cxa_atexit and __cxa_finalize implementation

Peter S. Mazinger ps.m at gmx.net
Mon Sep 26 18:30:03 UTC 2005


On Mon, 26 Sep 2005, Stephen Warren wrote:

> Peter S. Mazinger wrote:
> > On Sat, 24 Sep 2005, Stephen Warren wrote:
> > 
> > 
> >>Peter S. Mazinger wrote:
> >>
> >>>On Fri, 23 Sep 2005, Stephen Warren wrote:
> >>>
> >>>>I get this error:
> >>>>
> >>>>/opt/gcc-3.4.3-uclibc-0.9.27-i486/bin/i486-linux-uclibc-g++ \
> >>>>    -o dl-2 '-DDSO_NAME="./dso-2.so"' dl.cpp -ldl
> >>>>
> >>>>/mnt/severn_home_nobackup/swarren/gcc-3.4.3-uclibc-0.9.27-i486/bin/../lib/gcc/i486-linux-uclibc/3.4.3/../../../../i486-linux-uclibc/lib/crt1.o(.text+0x20): 
> >>>>In function `_start':
> >>>>crt0.S: undefined reference to `__uClibc_start_main'
> >>>>collect2: ld returned 1 exit status
> >>>
> >>>Are you using 0.9.27 or current? 0.9.27 used crt0.S and 
> >>>__uClibc_start_main, but 0.9.28 uses crt1.S and __uClibc_main.
> >>>For sure I wouldn't do any 0.9.27 tests (and your tests results won't be 
> >>>relevant for acceptance into svn), too many things changed.
> >>>For sure your linking stage is screwed up.
> >>
> >>I'm running the latest SVN version of uClibc.
> > 
> > 
> > I have asked, because above the name is 'gcc-3.4.3-uclibc-0.9.27-i486'
> > Something is mixed up, because it tries to use the older crt file 
> > (that asks for __uClibc_start_main) but the newer libc.
> 
> Oh. I think that's just because I copied the buildroot .config file from 
> the old buildroot tree into the svn buildroot tree, and the .config file 
> hard-codes the "staging directory" to the old name we use.
> 
> That said, I might not have deleted the content in that directory first, 
> so that could have been confusing buildroot/uclibc build in that run...

I have committed your patch and mine (with an additional compatibility 
option) to svn. Thanks.

Peter

-- 
Peter S. Mazinger <ps dot m at gmx dot net>           ID: 0xA5F059F2
Key fingerprint = 92A4 31E1 56BC 3D5A 2D08  BB6E C389 975E A5F0 59F2




More information about the uClibc mailing list