Bryan Kadzban wrote:
Bruce Dubbs wrote:
In your case though, you would only need to change to
--with-glibc-version=2.10, but I am guessing it would work with 2.11
anyway.
Yeah, I misunderstood the effect of the configure option (in part because of
the changes being discussed, to set the minimum *host* glibc requirement). I
thought it worked like glibc's --enable-kernel=xxx flag, in that the resulting
gcc would only work if the libc that *it* linked to (in its main gcc binary)
was at least the specified version. (Like libc.so abort()s at startup if the
kernel version isn't at least whatever glibc was configured for.) I figured
it would do something with symbol versioning or whatever to accomplish this.
Instead, according to:
http://gcc.gnu.org/ml/gcc-patches/2013-11/msg00619.html
the configure flag seems to be the minimum that the compiler will *target*,
i.e. it will only affect the compiler's output, not the compiler binary's
running, and actually only affects configure-time checks in the gcc build.
So I'm fine with that. --with-glibc-version=2.11 should indeed work on my
system. And I actually think changing the minimum host glibc back to whatever
it was (2.5.something?) might be a good idea, to avoid incorrect assumptions
in the future? But up to you I suppose.
I'm going to leave it at 2.11 for now. A 5 year old package should be enough.
Users always have the option of using a livecd with more recent package versions.
I'll reconsider if we start getting users with issues due to the change.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
