On Thursday April 5 2007 16:03, Kevin Day wrote:
> I have noticed segfaults registering in their tests at uclibc:8010
> as well as experienced segfaults
>
> The updates in the book also suggests this.
>
> I occasionally test their svn, but I am testing with linux-2.6.18.8
> and its headers.
> Perhaps these segfaults are fixed by using 2.6.20 kernel and headers?
>
> My last test was quite recent, but not this exact date, so maybe those
> particular segfaults I was experiencing are resolved..

Programs built with -fstack-protector-all and linked to libuClibc are 
segfaulting. I don't know why yet, but the segfaults are coming from 
libUclibc.so.

-fstack-protector (without the -all) works.

Using libssp works. I'm investigating the differences between GCC's ssp*.c and 
uClibc's. I have a couple things I want to check before giving up. If all 
else fails then I'll get GCC to link to libssp.so for uClibc. I'll have 
something safely working before the end of the night.

- Off topic -
There are also a couple more patches to add to GCC, but I doubt they're 
related to these segfaults so I'm not adding them yet. The uClibc.org GCC 
c99-snprintf patch (an Sed will do this), and the fix_fastmath (I want 
gcc-cvs urls for this one).

H.J. Lu fixed Binutils PR4304, so '--warn-shared-textrel --fatal-warnings' now 
works. uClibc can do this at run time (give a fatal error if a shared object 
contains text relocations), but now it can be hardcoded into ld(1), or in GCC 
specs, to work at compile time so Glibc can use it too.

'--warn-shared-textrel --fatal-warnings' can be placed beside '-pie" in the 
link specs, however this will also make mktemp(3) warnings fatal. ld's source 
can be manipulated to make warn-shared-textrel default to being fatal, while 
not affecting other warnings. Each has its own advantages. Making all linker 
warnings fatal has the advantage of forcing us to go out of our way to use 
mktemp(3), but I don't know what else is affected.

robert

Attachment: pgpdQExqfUnVH.pgp
Description: PGP signature

-- 
http://linuxfromscratch.org/mailman/listinfo/hlfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to