compiled with -fno-stack-protector and -U_FORTIFY_SOURCE

please do a test build on all architectures including community ports and report back about regressions/progressions.

Report on test output for all architectures. I think the differences are very minimal, and I would like to get the stack and fortify options back to the Ubuntu defaults.

Patch to reenable protections...

> I think the differences are very minimal, and I would like to get the stack
> and fortify options back to the Ubuntu defaults.

these are regressions. I don't think regressions are minimal.

proposing to unset the milestone until the regressions are fixed.

the regressions with openjdk-6-6b16~pre1 on i386 are:

FAILED: com/sun/jdi/BadHandshakeTest.java
FAILED: javax/management/remote/mandatory/notif/NotificationBufferDeadlockTest.java
FAILED: sun/jvmstat/monitor/MonitoredVm/MonitorVmStartTerminate.sh

Updated patch...

javax/management/remote/mandatory/notif/NotificationBufferDeadlockTest.java fails with stock local version too:

Testing protocol jmxmp
TEST FAILED: GOT EXCEPTION:
java.security.AccessControlException: access denied (javax.management.MBeanTrustPermission register)
        at java.security.AccessControlContext.checkPermission(AccessControlContext.java:342)
        at java.lang.SecurityManager.checkPermission(SecurityManager.java:585)
        at com.sun.jmx.interceptor.DefaultMBeanServerInterceptor.checkMBeanTrustPermission(DefaultMBeanServerInterceptor.java:1868)
        at com.sun.jmx.interceptor.DefaultMBeanServerInterceptor.registerMBean(DefaultMBeanServerInterceptor.java:328)
        at com.sun.jmx.mbeanserver.JmxMBeanServer.registerMBean(JmxMBeanServer.java:516)
        at NotificationBufferDeadlockTest.test(NotificationBufferDeadlockTest.java:116)
        at NotificationBufferDeadlockTest.main(NotificationBufferDeadlockTest.java:98)
        at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
        at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
        at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
        at java.lang.reflect.Method.invoke(Method.java:616)
        at com.sun.javatest.regtest.MainAction$SameVMThread.run(MainAction.java:595)
        at java.lang.Thread.run(Thread.java:636)

sun/jvmstat/monitor/MonitoredVm/MonitorVmStartTerminate.sh did not fail in current build tests.

com/sun/jdi/BadHandshakeTest.java is a "real" problem, though:

*** buffer overflow detected ***: /usr/lib/jvm/java-6-openjdk/jre/bin/java terminated
======= Backtrace: =========
/lib/libc.so.6(__fortify_fail+0x40)[0xf7e99a90]
/lib/libc.so.6[0xf7e98aa0]
/lib/libc.so.6[0xf7e97dca]
/usr/lib/jvm/java-6-openjdk/jre/lib/i386/libdt_socket.so[0xf70f6eb7]
/usr/lib/jvm/java-6-openjdk/jre/lib/i386/libdt_socket.so[0xf70f7066]
/usr/lib/jvm/java-6-openjdk/jre/lib/i386/libjdwp.so[0xf7128357]
/usr/lib/jvm/java-6-openjdk/jre/lib/i386/server/libjvm.so[0xf760af14]
/usr/lib/jvm/java-6-openjdk/jre/lib/i386/server/libjvm.so[0xf760affc]
/usr/lib/jvm/java-6-openjdk/jre/lib/i386/server/libjvm.so[0xf779c2cc]
/usr/lib/jvm/java-6-openjdk/jre/lib/i386/server/libjvm.so[0xf779c38a]
/usr/lib/jvm/java-6-openjdk/jre/lib/i386/server/libjvm.so[0xf76c0931]
/lib/libpthread.so.0[0xf7f20680]
/lib/libc.so.6(clone+0x5e)[0xf7e84b4e]
======= Memory map: ========
08048000-08051000 r-xp 00000000 fc:1d 12097 /usr/lib/jvm/java-6-openjdk/jre/bin/java
08051000-08052000 r--p 00008000 fc:1d 12097 /usr/lib/jvm/java-6-openjdk/jre/bin/java
08052000-08053000 rw-p 00009000 fc:1d 12097 /usr/lib/jvm/java-6-openjdk/jre/bin/java
08484000-0854e000 rw-p 00000000 00:00 0 [heap]
ab494000-ab4be000 r-xp 00000000 fc:1d 65590 /lib/libgcc_s.so.1
ab4be000-ab4bf000 r--p 00029000 fc:1d 65590 /lib/libgcc_s.so.1
ab4bf000-ab4c0000 rw-p 0002a000 fc:1d 65590 /lib/libgcc_s.so.1
ab4c0000-ab4c3000 ---p 00000000 00:00 0
ab4c3000-ab510000 rw-p 00000000 00:00 0
ab510000-ab513000 ---p 00000000 00:00 0
ab513000-ab560000 rw-p 00000000 00:00 0
ab560000-ab563000 ---p 00000000 00:00 0
ab563000-ab5b0000 rw-p 00000000 00:00 0
ab5b0000-ab5b3000 ---p 00000000 00:00 0
ab5b3000-ab600000 rw-p 00000000 00:00 0
ab600000-ab621000 rw-p 00000000 00:00 0
ab621000-ab700000 ---p 00000000 00:00 0
ab720000-ab723000 ---p 00000000 00:00 0
ab723000-ab770000 rw-p 00000000 00:00 0
ab770000-ab771000 ---p 00000000 00:00 0
ab771000-ab823000 rw-p 00000000 00:00 0
ab823000-ab9b7000 r--s 038e2000 fc:1d 12095 /usr/lib/jvm/java-6-openjdk/jre/lib/rt.jar
ab9b7000-ab9b8000 ---p 00000000 00:00 0
ab9b8000-aba37000 rw-p 00000000 00:00 0
aba37000-aba38000 ---p 00000000 00:00 0
aba38000-abab7000 rw-p 00000000 00:00 0
abab7000-abab8000 ---p 00000000 00:00 0
abab8000-abb37000 rw-p 00000000 00:00 0
abb37000-abb38000 ---p 00000000 00:00 0
abb38000-abbb7000 rw-p 00000000 00:00 0
abbb7000-abbbf000 rwxp 00000000 00:00 0
abbbf000-abbf7000 rwxp 00000000 00:00 0
abbf7000-abc2f000 rwxp 00000000 00:00 0
abc2f000-abdbf000 rwxp 00000000 00:00 0
abdbf000-abdc7000 rwxp 00000000 00:00 0
abdc7000-abdff000 rwxp 00000000 00:00 0
abdff000-abe37000 rwxp 00000000 00:00 0
abe37000-abfc6000 rwxp 00000000 00:00 0
abfc6000-abfcd000 rwxp 00000000 00:00 0
abfcd000-abfff000 rwxp 00000000 00:00 0
abfff000-ad000000 rwxp 00000000 00:00 0
ad000000-b4000000 rwxp 00000000 00:00 0
b4000000-bae10000 rwxp 00000000 00:00 0
bae10000-ece40000...

Given this is a singular specific bug (not a java-wide problem), I'd like to push to get these options enabled for Feature Freeze, along with bug 409736.

Bug 419018 filed and solved.

Current testsuite differences between 6b16-1.6~pre1-0ubuntu1 and my most recent build with fixes for 330713, 409736, and 419018 seems to introduce one progression and no regressions:

-FAILED: sun/security/ssl/javax/net/ssl/NewAPIs/SessionCacheSizeTests.java
+Passed: sun/security/ssl/javax/net/ssl/NewAPIs/SessionCacheSizeTests.java

This bug was fixed in the package openjdk-6 - 6b16-1.6~pre2-0ubuntu1

---------------
openjdk-6 (6b16-1.6~pre2-0ubuntu1) karmic; urgency=low

  * Update IcedTea from the 1.6 release branch:
    - Fix buffer overflow in debugger's socket handler (Kees Cook).
      https://bugs.openjdk.java.net/show_bug.cgi?id=100103. LP: #409736.
    - plugin fixes.
  * Move the pulseaudio recommendation to a suggestion, don't build-depend
    on pulseaudio.
  * Build for armv6 (on armel).

  [ Kees Cook ]
  * debian/rules: Re-enable fortification and stack protector
    (LP: #330713).
  * Adding stack markings to the x86 assembly for not using executable
    stack. LP: #419018.

 -- Matthias Klose <email address hidden> Fri, 28 Aug 2009 18:51:34 +0200