| Message ID | 1346356090-13811-1-git-send-email-sgw@linux.intel.com |
|---|---|
| State | New |
| Headers | show |
diff --git a/meta-yocto/conf/distro/poky-tiny.conf b/meta-yocto/conf/distro/poky-tiny.conf index 8ae1d85..bfb4432 100644 --- a/meta-yocto/conf/distro/poky-tiny.conf +++ b/meta-yocto/conf/distro/poky-tiny.conf @@ -63,6 +63,7 @@ ASSUME_PROVIDED += "pkgconfig$" # Reconfigure eglibc for a smaller installation # Comment out any of the lines below to disable them in the build DISTRO_FEATURES_LIBC_TINY = "libc-libm libc-crypt" +DISTRO_FEATURES_LIBC_TINY_append_x86-64 = " libc-libm-big" # Required for "who" DISTRO_FEATURES_LIBC_MINIMAL = "libc-utmp libc-getlogin"
On Thu, Aug 30, 2012 at 12:48 PM, Saul Wold <sgw@linux.intel.com> wrote: > The libc-libm-big option allows eglibc to correctly configure > itself to include the correct math functions for 64bit operations > > [YOCTO #2943] > > Signed-off-by: Saul Wold <sgw@linux.intel.com> I'm curious, is this needed for other 64 bit platforms, e.g. powerpc64?
On Thu, 2012-08-30 at 13:04 -0700, Chris Larson wrote: > On Thu, Aug 30, 2012 at 12:48 PM, Saul Wold <sgw@linux.intel.com> wrote: > > The libc-libm-big option allows eglibc to correctly configure > > itself to include the correct math functions for 64bit operations > > > > [YOCTO #2943] > > > > Signed-off-by: Saul Wold <sgw@linux.intel.com> > > I'm curious, is this needed for other 64 bit platforms, e.g. powerpc64? Also out of curiosity, what's the use-case for 64-bit userspace in a "tiny" context? I would have thought that you'd generally want x32. p.
On Thu, Aug 30, 2012 at 1:13 PM, Phil Blundell <philb@gnu.org> wrote: > On Thu, 2012-08-30 at 13:04 -0700, Chris Larson wrote: >> On Thu, Aug 30, 2012 at 12:48 PM, Saul Wold <sgw@linux.intel.com> wrote: >> > The libc-libm-big option allows eglibc to correctly configure >> > itself to include the correct math functions for 64bit operations >> > >> > [YOCTO #2943] >> > >> > Signed-off-by: Saul Wold <sgw@linux.intel.com> >> >> I'm curious, is this needed for other 64 bit platforms, e.g. powerpc64? > > Also out of curiosity, what's the use-case for 64-bit userspace in a > "tiny" context? I would have thought that you'd generally want x32. > I guess its for sake of completeness. > p. > > > > _______________________________________________ > Openembedded-core mailing list > Openembedded-core@lists.openembedded.org > http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
On Thu, Aug 30, 2012 at 12:48 PM, Saul Wold <sgw@linux.intel.com> wrote: > The libc-libm-big option allows eglibc to correctly configure > itself to include the correct math functions for 64bit operations > > [YOCTO #2943] > > Signed-off-by: Saul Wold <sgw@linux.intel.com> > --- > meta-yocto/conf/distro/poky-tiny.conf | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/meta-yocto/conf/distro/poky-tiny.conf b/meta-yocto/conf/distro/poky-tiny.conf > index 8ae1d85..bfb4432 100644 > --- a/meta-yocto/conf/distro/poky-tiny.conf > +++ b/meta-yocto/conf/distro/poky-tiny.conf > @@ -63,6 +63,7 @@ ASSUME_PROVIDED += "pkgconfig$" > # Reconfigure eglibc for a smaller installation > # Comment out any of the lines below to disable them in the build > DISTRO_FEATURES_LIBC_TINY = "libc-libm libc-crypt" > +DISTRO_FEATURES_LIBC_TINY_append_x86-64 = " libc-libm-big" > > # Required for "who" > DISTRO_FEATURES_LIBC_MINIMAL = "libc-utmp libc-getlogin" there is a bug in eglibc if it cant build without libc-libm-big it should have not caused any undefined symbol failures because all it should have done is not use double precision so your computations will be less accurate. > -- > 1.7.9.5 > > > _______________________________________________ > Openembedded-core mailing list > Openembedded-core@lists.openembedded.org > http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
On 08/30/2012 02:49 PM, Khem Raj wrote: > On Thu, Aug 30, 2012 at 12:48 PM, Saul Wold <sgw@linux.intel.com> wrote: >> The libc-libm-big option allows eglibc to correctly configure >> itself to include the correct math functions for 64bit operations >> >> [YOCTO #2943] >> >> Signed-off-by: Saul Wold <sgw@linux.intel.com> >> --- >> meta-yocto/conf/distro/poky-tiny.conf | 1 + >> 1 file changed, 1 insertion(+) >> >> diff --git a/meta-yocto/conf/distro/poky-tiny.conf b/meta-yocto/conf/distro/poky-tiny.conf >> index 8ae1d85..bfb4432 100644 >> --- a/meta-yocto/conf/distro/poky-tiny.conf >> +++ b/meta-yocto/conf/distro/poky-tiny.conf >> @@ -63,6 +63,7 @@ ASSUME_PROVIDED += "pkgconfig$" >> # Reconfigure eglibc for a smaller installation >> # Comment out any of the lines below to disable them in the build >> DISTRO_FEATURES_LIBC_TINY = "libc-libm libc-crypt" >> +DISTRO_FEATURES_LIBC_TINY_append_x86-64 = " libc-libm-big" >> >> # Required for "who" >> DISTRO_FEATURES_LIBC_MINIMAL = "libc-utmp libc-getlogin" > > > there is a bug in eglibc if it cant build without libc-libm-big > it should have not caused any undefined symbol failures because > all it should have done is not use double precision so your computations > will be less accurate. > Khem, You are correct the problem is that if the libm-big option is not specificed it grabs the dbl-wrap files for some sysdep function, but still uses the dbl-64 functions for others. Should I file an eglibc bug? Sau! >> -- >> 1.7.9.5 >> >> >> _______________________________________________ >> Openembedded-core mailing list >> Openembedded-core@lists.openembedded.org >> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core > >
On Fri, Aug 31, 2012 at 9:31 AM, Saul Wold <sgw@linux.intel.com> wrote: > > Should I file an eglibc bug? yes
On 08/30/2012 02:49 PM, Khem Raj wrote: > On Thu, Aug 30, 2012 at 12:48 PM, Saul Wold <sgw@linux.intel.com> wrote: >> The libc-libm-big option allows eglibc to correctly configure >> itself to include the correct math functions for 64bit operations >> >> [YOCTO #2943] >> >> Signed-off-by: Saul Wold <sgw@linux.intel.com> >> --- >> meta-yocto/conf/distro/poky-tiny.conf | 1 + >> 1 file changed, 1 insertion(+) >> >> diff --git a/meta-yocto/conf/distro/poky-tiny.conf b/meta-yocto/conf/distro/poky-tiny.conf >> index 8ae1d85..bfb4432 100644 >> --- a/meta-yocto/conf/distro/poky-tiny.conf >> +++ b/meta-yocto/conf/distro/poky-tiny.conf >> @@ -63,6 +63,7 @@ ASSUME_PROVIDED += "pkgconfig$" >> # Reconfigure eglibc for a smaller installation >> # Comment out any of the lines below to disable them in the build >> DISTRO_FEATURES_LIBC_TINY = "libc-libm libc-crypt" >> +DISTRO_FEATURES_LIBC_TINY_append_x86-64 = " libc-libm-big" >> >> # Required for "who" >> DISTRO_FEATURES_LIBC_MINIMAL = "libc-utmp libc-getlogin" > > > there is a bug in eglibc if it cant build without libc-libm-big > it should have not caused any undefined symbol failures because > all it should have done is not use double precision so your computations > will be less accurate. Agreed. ~200K is significant when we are talking about < 4MB images. However, so long as we track the eglibc bug, I am fine with including this workaround to get things going. I would ask that a comment be added which calls this out as a workaround for the eglibc bug (a bug number would be great). Thanks, Darren > >> -- >> 1.7.9.5 >> >> >> _______________________________________________ >> Openembedded-core mailing list >> Openembedded-core@lists.openembedded.org >> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core > > _______________________________________________ > Openembedded-core mailing list > Openembedded-core@lists.openembedded.org > http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core >
The libc-libm-big option allows eglibc to correctly configure itself to include the correct math functions for 64bit operations [YOCTO #2943] Signed-off-by: Saul Wold <sgw@linux.intel.com> --- meta-yocto/conf/distro/poky-tiny.conf | 1 + 1 file changed, 1 insertion(+)