x86_64-pc-linux-gnu-gcc -DHAVE_CONFIG_H -I. -I.. -I.. -I../include -I../VEX/pub -DVGA_amd64=1 -DVGO_linux=1 -DVGP_amd64_linux=1 -I../coregrind -DVG_LIBDIR="\"/usr/lib64/valgrind"\" -DVG_PLATFORM="\"amd64-linux\"" -m64 -fomit-frame-pointer -fno-stack-protector -O2 -g -Wall -Wmissing-prototypes -Wshadow -Wpointer-arith -Wstrict-prototypes -Wmissing-declarations -Wno-format-zero-length -fno-strict-aliasing -Wno-long-long -march=native -O2 -Wno-pointer-sign -fno-stack-protector -MT libcoregrind_amd64_linux_a-m_signals.o -MD -MP -MF .deps/libcoregrind_amd64_linux_a-m_signals.Tpo -c -o libcoregrind_amd64_linux_a-m_signals.o `test -f 'm_signals.c' || echo './'`m_signals.c Imho should be dropped: -fomit-frame-pointer -O2 -g
I just added 3.7.0 to the tree but did not address this issue. Can you test removing these flags and see if its okay? I would think it should be but valgrind is an unusual package.
there are still same flags that I said
Yes, I'm asking you to remove the yourself and test whether valgrind breaks before I remove them in the ebuild. I said this very clearly in my previous post with explanation.
This is not a package where user defined cflags work well, eg bug #335065. I'm not going to touch any of these flags, including -g, unless there is ample testing that valgrind doesn't break in the process.
Here's other examples: bug #403827 and another bug #335065