[Bug go/48122] New: crypto/aes test fails on 32-bit Solaris 11/x86

[Bug go/48122] New: crypto/aes test fails on 32-bit Solaris 11/x86

[ro at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48122

           Summary: crypto/aes test fails on 32-bit Solaris 11/x86
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: go
        AssignedTo: ian@airs.com
        ReportedBy: ro@gcc.gnu.org
              Host: i386-pc-solaris2.11
            Target: i386-pc-solaris2.11
             Build: i386-pc-solaris2.11


The crypto/aes test recently started failing on 32-bit Solaris 11/x86 (only,
64-bit is fine):

FAIL: crypto/aes
--- FAIL: aes.TestCipherDecrypt
    decryptBlock 0: plain[0] = 0x0, want 0x32
    decryptBlock 1: plain[1] = 0x0, want 0x11
    decryptBlock 2: plain[1] = 0x0, want 0x11
    decryptBlock 3: plain[1] = 0x0, want 0x11
--- FAIL: aes.TestCipherEncrypt
    Cipher.Encrypt 0: out[0] = 0x0, want 0x32
    Cipher.Encrypt 1: out[1] = 0x0, want 0x11
    Cipher.Encrypt 2: out[1] = 0x0, want 0x11
    Cipher.Encrypt 3: out[1] = 0x0, want 0x11
FAIL
make: *** [crypto/aes/check] Error 1

I'm pretty sure it worked before, but cannot be certain since libgo testsuite
results don't show up in mail-report.log.

[Bug go/48122] crypto/aes test fails on 32-bit Solaris 11/x86

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48122

--- Comment #1 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2011-04-11 13:15:45 UTC ---
I noticed that this error only happens in a modifed tree with the libgo
multilib check patch

    http://gcc.gnu.org/ml/gcc-patches/2011-04/msg00247.html

Investigation showed that _testmain.go differs between the two trees:

--- ./_testmain.go      2011-04-09 20:39:58.824574614 +0200
+++
/var/gcc/gcc-4.7.0-20110408/11-gcc-gas/i386-pc-solaris2.11/libgo/crypto-aes-check/_testmain.go
     2011-04-09 20:38:06.328947345 +0200
@@ -14,8 +14,8 @@
        {"aes.TestExpandKey", aes.TestExpandKey},
        {"aes.TestEncryptBlock", aes.TestEncryptBlock},
        {"aes.TestDecryptBlock", aes.TestDecryptBlock},
-       {"aes.TestCipherEncrypt", aes.TestCipherEncrypt},
        {"aes.TestCipherDecrypt", aes.TestCipherDecrypt},
+       {"aes.TestCipherEncrypt", aes.TestCipherEncrypt},
 }
 var benchmarks = []testing.InternalBenchmark{ //
 }

It turns out that this was caused by passing $GOCFLAGS in CHECK (GC),
which was meant to pass the multilib flags, but isn't necessary since
those are included in $GOC.

But unfortunately the default (-g -O2) changes the outcome of this and a
couple of others tests.  I don't think the tests should be so fragile.

    Rainer

[Bug go/48122] crypto/aes test fails on 32-bit Solaris 11/x86

[ian at airs dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48122

Ian Lance Taylor <ian at airs dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED

--- Comment #2 from Ian Lance Taylor <ian at airs dot com> 2012-02-14 19:47:16 UTC ---
I believe this is fixed now.  The testsuite compilation now uses
-fno-toplevel-reorder to ensure that the tests are run in the expected order.

[Bug go/48122] crypto/aes test fails on 32-bit Solaris 11/x86

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48122

--- Comment #3 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2012-02-16 19:23:05 UTC ---
> --- Comment #2 from Ian Lance Taylor <ian at airs dot com> 2012-02-14 19:47:16 UTC ---
> I believe this is fixed now.  The testsuite compilation now uses
> -fno-toplevel-reorder to ensure that the tests are run in the expected order.

True, I just neglected to check and close those old PRs.

    Rainer