Age | Commit message (Collapse) | Author |
|
target_includedir
This fixes the problem where gcc fails to detect presence of libssp
and does not detect powerpc long double 128 bit support correctly on systems which do
not use common <sysroot>/usr/include paths like we have in micro distro
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Acked-by: Tom Rini <tom_rini@mentor.com>
|
|
This fixes the "libstdc++ isn't installed into my rootfs anymore" type of problems after a full rebuild of all packages
This bumps PR of gcc to force a rebuild, but that is *NOT* enough to fix the packages built before this fix, those still need rebuilding
Acked-by: Khem Raj <raj.khem@gmail.com>
Acked-by: Dallas Foley <dfoley@telus.net>
Acked-by: Graeme Gregory <dp@xora.org.uk>
Signed-off-by: Koen Kooi <koen@openembedded.org>
|
|
make that happen
* The existing patch did redefine the SHLIB_MAKE_SOLINK and
SHLIB_INSTALL_SOLINK but it was ineffective because the original
definitions appeared after it. We needed to rearrange the order
* t-slibgcc-libgcc is already there so no need to patch it.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
removed unneeded FILESPATHPKG
tested by running bitbake -c configure -b on all
gcc*_4.4*.bb recipes
Signed-off-by: Frans Meulenbroeks <fransmeulenbroeks@gmail.com>
Acked-by: Khem Raj <raj.khem@gmail.com> (on irc)
|
|
When compiling gcc target code make passes -I(GMPINC)
but this one points to the host dir (it is set to HOST_GMPINC
which in oe points to tmp/sysroots/i686-linux/usr/include).
This patch breaks the env var $(INCLUDES) in two and
only uses the TARGET specific part for compiling libgcc
and the crt stuff.
The patch has been verified by baking all affected recipes with
bitbake -c configure -b recipe
Signed-off-by: Frans Meulenbroeks <fransmeulenbroeks@gmail.com>
|
|
We don't need these paths to be set and they further break things if
we use --sysroot rather than -isystem
Signed-off-by: Tom Rini <tom_rini@mentor.com>
|
|
breaking ABI for armv7
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
* ARMv7 was using -mtp=soft where as the CP15 register for TLS
is available and should be used. This should improve the performance
of TLS access.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
Target CFLAGS (LIBCFLAGS in fact) are leaking into native parts of the
build. Fix that (thus making gcc buildable for ppc, e.g.).
Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
|
|
For .diff/.patch you need to apply manually, you can specify apply=no.
Signed-off-by: Chris Larson <chris_larson@mentor.com>
Acked-by: Denys Dmytriyenko <denis@denix.org>
Acked-by: Khem Raj <raj.khem@gmail.com>
|
|
I think this makes the behavior rather more clear.
Signed-off-by: Chris Larson <chris_larson@mentor.com>
Acked-by: Denys Dmytriyenko <denis@denix.org>
Acked-by: Khem Raj <raj.khem@gmail.com>
|
|
* For uclibc __to_type is typedef'ed to const __ctype_touplow_t*
which is a short int. So the mangled alias breaks because on
glibc __to_type is int* and mangled alias is set accordingly
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
* This change converts libgcc_s.so to a linker script so when -lgcc_s is
used then it actually links to a group of libraries and one of those
libraries is libgcc. This fixes the problem of requiring to link with
libgcc _always_ on ppc to get the out-of-line epilogue/prologue
functions.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
* Reset INC_PR and update checksums.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|