On Tue, Feb 06, 2018 at 11:14:08AM +0000, Tsimbalist, Igor V wrote:
> Coincidentally, I have worked on the same patch. Please look at the patch, I
> uploaded it to the bug. The main differences are
>
> - updated the output messages to be more informative;
> - updated the tests and add couple of new tests to check the messages;
> - fixed a typo in the doc file related to fcf-protection;
>
> I am ok with the changes in i386.c but would like to update the messages.
> Could you incorporate my changes and proceed? Or would you like me to finish
> the fix?
The r257414 change broke bootstrap on both x86_64-linux and i686-linux,
in both the bootstrap ICEs during libitm build of aatree.cc:
/.../gcc/obj70/./gcc/xg++ -B/.../gcc/obj70/./gcc/ -nostdinc++ -nostdinc++
-I/.../gcc/obj70/x86_64-pc-linux-gnu/libstdc++-v3/include/x86_64-pc-linux-gnu
-I/.../gcc/obj70/x86_64-pc-linux-gnu/libstdc++-v3/include
-I/.../gcc/libstdc++-v3/libsupc++ -I/.../gcc/libstdc++-v3/include/backward
-I/.../gcc/libstdc++-v3/testsuite/util
-L/.../gcc/obj70/x86_64-pc-linux-gnu/libstdc++-v3/src
-L/.../gcc/obj70/x86_64-pc-linux-gnu/libstdc++-v3/src/.libs
-L/.../gcc/obj70/x86_64-pc-linux-gnu/libstdc++-v3/libsupc++/.libs
-B/.../gcc/obj70/x86_64-pc-linux-gnu/libstdc++-v3/src/.libs
-B/.../gcc/obj70/x86_64-pc-linux-gnu/libstdc++-v3/libsupc++/.libs
-B/usr/local/x86_64-pc-linux-gnu/bin/ -B/usr/local/x86_64-pc-linux-gnu/lib/
-isystem /usr/local/x86_64-pc-linux-gnu/include -isystem
/usr/local/x86_64-pc-linux-gnu/sys-include -DHAVE_CONFIG_H -I.
-I../../../libitm -I../../../libitm/config/linux/x86
-I../../../libitm/config/linux -I../../../libitm/config/x86
-I../../../libitm/config/posix -I../../../libitm/config/generic
-I../../../libitm -mrtm -Wall -pthread -Werror -fcf-protection -mcet
-std=gnu++0x -funwind-tables -fno-exceptions -fno-rtti -fabi-version=4 -g -O2
-D_GNU_SOURCE -MT aatree.lo -MD -MP -MF .deps/aatree.Tpo -c
../../../libitm/aatree.cc -fPIC -DPIC -o .libs/aatree.o
/.../gcc/obj70/gcc/include/ia32intrin.h:56:28: internal compiler error: in
ix86_option_override_internal, at config/i386/i386.c:4948
The problem is that the enum has
{CF_NONE, CF_BRANCH, CF_RETURN, CF_FULL, CF_SET}
values and the hook does:
opts->x_flag_cf_protection =
(cf_protection_level) (opts->x_flag_cf_protection | CF_SET);
after the switch (note, bad formatting), and if called again, we end up with
e.g. CF_FULL | CF_SET value which the switch doesn't handle.
It isn't clear what you want to do, either ignore the CF_SET | ...
values in the switch instead of gcc_unreachable on them, or
switch (flag_cf_protection & ~CF_SET)
or something similar.
Jakub
