Same thing with gcc 4.9.0-5:

 http://arm.koji.fedoraproject.org/koji/buildinfo?buildID=195103


[ 35%] Building CXX object sql/CMakeFiles/binlog.dir/log_event.cc.o
cd /builddir/build/BUILD/mysql-5.6.17/build/sql && /usr/bin/c++   -DHAVE_CONFIG_H -DHAVE_EVENT_SCHEDULER -DHAVE_OPENSSL -DMYSQL_SERVER -O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fPIC -Wall -Wno-unused-parameter -O3 -g -fabi-version=2 -fno-omit-frame-pointer -fno-strict-aliasing -DDBUG_OFF -DMY_PTHREAD_FASTMUTEX=1 -I/builddir/build/BUILD/mysql-5.6.17/build/include -I/builddir/build/BUILD/mysql-5.6.17/include -I/builddir/build/BUILD/mysql-5.6.17/sql -I/builddir/build/BUILD/mysql-5.6.17/regex -I/builddir/build/BUILD/mysql-5.6.17/build/sql    -o CMakeFiles/binlog.dir/log_event.cc.o -c /builddir/build/BUILD/mysql-5.6.17/sql/log_event.cc
/builddir/build/BUILD/mysql-5.6.17/storage/perfschema/pfs_account.cc: In function 'PFS_account* find_or_create_account(PFS_thread*, const char*, uint, const char*, uint)':
/builddir/build/BUILD/mysql-5.6.17/storage/perfschema/pfs_account.cc:320:1: error: could not split insn
 }
 ^
(insn 474 1814 1392 (parallel [
            (set (reg:SI 0 x0 [orig:162 D.17087 ] [162])
                (mem/v:SI (reg/f:DI 1 x1 [orig:117 D.17094 ] [117]) [-1  S4 A32]))
            (set (mem/v:SI (reg/f:DI 1 x1 [orig:117 D.17094 ] [117]) [-1  S4 A32])
                (unspec_volatile:SI [
                        (ior:SI (mem/v:SI (reg/f:DI 1 x1 [orig:117 D.17094 ] [117]) [-1  S4 A32])
                            (const_int 0 [0]))
                        (const_int 5 [0x5])
                    ] UNSPECV_ATOMIC_OP))
            (clobber (reg:CC 66 cc))
            (clobber (reg:SI 2 x2 [335]))
            (clobber (reg:SI 4 x4 [336]))
        ]) /builddir/build/BUILD/mysql-5.6.17/include/my_atomic.h:217 2197 {atomic_fetch_orsi}
     (expr_list:REG_UNUSED (reg:CC 66 cc)
        (expr_list:REG_UNUSED (reg:SI 4 x4 [336])
            (expr_list:REG_UNUSED (reg:SI 2 x2 [335])
                (nil)))))
/builddir/build/BUILD/mysql-5.6.17/storage/perfschema/pfs_account.cc:320:1: internal compiler error: in final_scan_insn, at final.c:2952
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://bugzilla.redhat.com/bugzilla> for instructions.
{standard input}: Assembler messages:
{standard input}: Error: open CFI at the end of file; missing .cfi_endproc directive

Anything I can do to help this forward?

Created attachment 909284 [details]
Preprocessed source

Preprocessed source.

Got a Foundation Model based system up and running. 

Did some digging.

Attachment in C#3 is preprocessed source as saved in /tmp.

When using -O1 (and lower) or adding __attribute__((optimize (1))) to
problematic function:

--- storage/perfschema/pfs_account.cc.orig
+++ storage/perfschema/pfs_account.cc
@@ -204,6 +204,11 @@
 PFS_account *
 find_or_create_account(PFS_thread *thread,
                          const char *username, uint username_length,
+                         const char *hostname, uint hostname_length) __attribute__((optimize (1)));
+
+PFS_account *
+find_or_create_account(PFS_thread *thread,
+                         const char *username, uint username_length,
                          const char *hostname, uint hostname_length)
 {
   if (account_max == 0)

things works.

To reproduce:

install build req:

 yum install gcc gcc-c++ cmake ncurses-devel openssl-devel libedit-devel libaio-devel

Download mysql tarball:

 wget http://cdn.mysql.com/Downloads/MySQL-5.6/mysql-5.6.19.tar.gz

Unpack, create dir and run cmake:

 tar xzvf mysql-5.6.19.tar.gz
 cd mysql-5.6.19
 mkdir build && cd build
 cmake ..

To go storage/perfschema and try to compile pfs_account.cc file:

 cd storage/perfschema

Compile:

 /usr/bin/c++  \
 -DHAVE_CONFIG_H \
 -DMYSQL_SERVER \
 -fno-delete-null-pointer-checks \
 -fPIC -Wall -Wno-unused-parameter \
 -O2 -g -fabi-version=2  \
 -fno-omit-frame-pointer \
 -fno-strict-aliasing \
 -DDBUG_OFF -DMY_PTHREAD_FASTMUTEX=1 \
 -I../../../build/include \
 -I../../..   \
 -I../../../include \
 -I../../../sql  \
 -I../../../regex  \
 -I../../../extra/yassl/include \
 -o CMakeFiles/perfschema.dir/pfs_account.cc.o \
 -c ../../../storage/perfschema/pfs_account.cc

Anything more I can do?

Sorry for not responding, is this still an issue?

Good and bad news: 

 5.0.0-0.10.fc22 in coming Fedora 22 works 
 4.9.2-1.fc21 in Fedora 21 dont works

Logs (f22): http://arm.koji.fedoraproject.org/koji/taskinfo?taskID=2878458
Logs (f21): http://arm.koji.fedoraproject.org/koji/taskinfo?taskID=2878462

As we have a workaround for Fedora 21 that works:

 http://arm.koji.fedoraproject.org/koji/buildinfo?buildID=259526

its not very important to backport the fix to 4.9.

Hmm, 5.0.0-0.10 seems to trigger some tests to fail, which passes with 4.9.2 and this patch

 http://pkgs.fedoraproject.org/cgit/community-mysql.git/tree/community-mysql-5.6.19-gcc49-aarch64-opt.patch

Will retry with 5.0.0-0.10 with patch to check.

With 5.0.0-10 four tests (funcs_1.memory_func_view funcs_1.myisam_func_view main.bigint funcs_1.innodb_func_view) fails on all archs:

 http://arm.koji.fedoraproject.org/koji/taskinfo?taskID=2878633
 http://koji.fedoraproject.org/koji/taskinfo?taskID=8911833

while all tests passes with 4.9.2.

Seems like new bug introduced in gcc 5.0.0.

This bug appears to have been reported against 'rawhide' during the Fedora 22 development cycle.
Changing version to '22'.

More information and reason for this action is here:
https://fedoraproject.org/wiki/Fedora_Program_Management/HouseKeeping/Fedora22

By build gcc 5.0.1-0.1.fc23 issue has been resolved, no extra patches required to build, and %check passes. Thanks.