runtime: test failure on Solaris

[gopherbot]

by zheganin@enaza.ru:

runtime package fails to pass test. Actually, this is 100% reprodicible meaning it
fails, but reasons can differ each time (I just hit all.bash every time):
- can just silently "FAIL" without stating a reason, prior to
"GOMAXPROCS=2" phase
- can state "fatal error: unexpected signal during runtime execution" as a
reason on "GOMAXPROCS=2" phase
- can state "panic: runtime error: invalid memory address or nil pointer
dereference" as a reason

OS: Solaris 11.1 amd64
Compiler:

Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/gcc/4.5/lib/gcc/i386-pc-solaris2.11/4.5.2/lto-wrapper
Target: i386-pc-solaris2.11
Configured with:
/builds/hudson/workspace/nightly-update/build/i386/components/gcc45/gcc-4.5.2/configure
CC=/ws/on11update-tools/SUNWspro/sunstudio12.1/bin/cc
CXX=/ws/on11update-tools/SUNWspro/sunstudio12.1/bin/CC --prefix=/usr/gcc/4.5
--mandir=/usr/gcc/4.5/share/man --bindir=/usr/gcc/4.5/bin --libdir=/usr/gcc/4.5/lib
--sbindir=/usr/gcc/4.5/sbin --infodir=/usr/gcc/4.5/share/info
--libexecdir=/usr/gcc/4.5/lib --enable-languages=c,c++,fortran,objc --enable-shared
--with-gmp-include=/usr/include/gmp --with-mpfr-include=/usr/include/mpfr
--without-gnu-ld --with-ld=/usr/bin/ld --with-gnu-as --with-as=/usr/gnu/bin/as
CFLAGS='-g -O2 '
Thread model: posix
gcc version 4.5.2 (GCC)

Session logs attached.

Attachments:

1. session01.log (16448 bytes)
2. session02.log (34838 bytes)

[ianlancetaylor]

Comment 1:

Labels changed: added repo-main, release-go1.4, os-solaris.

[gopherbot]

Comment 2 by zheganin@enaza.ru:

Yeah, terribly sorry - forgot to mention version - this is 1.3.

[4ad]

Comment 3:

How many CPUs?
Possibly duplicate of issue #7554 and/or issue #7860.

[gopherbot]

Comment 4 by zheganin@enaza.ru:

16 logical.
[root@hyperion src]# mpstat 1 1
 CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw syscl  usr sys  wt idl
   0   58   0   19   921  383  965    1   55  641    1  4089    2   3   0  95
   1   47   0   10  2243 1891 1195    2   55  623    2  4437    2   5   0  93
   2   57   0   18   955  625  928    1   54  669    1  3182    2   3   0  95
   3   81   0   17   599  265  975    2   47  730    2  3903    2   3   0  95
   4   77   0   16   696  320  971    1   48  627    1  1634    1   2   0  97
   5   43   0   15   782  359 1158    2   59  711    1  1740    1   2   0  97
   6   74   0   17   724  334 1005    1   48  664    1  1689    1   2   0  97
   7   72   0   21   772  355 1121    1   58  756    1  1866    1   2   0  97
   8   59   0   17   617  266 1002    2   57  712    2  3739    2   3   0  95
   9   62   0   15   508  206  789    1   42  542    1  3451    2   3   0  96
  10   59   0   18   636  277 1009    1   57  732    2  3620    2   3   0  95
  11   66   0   15   515  220  802    1   41  576    1  3218    2   2   0  96
  12   82   0   19   735  341 1054    1   50  713    1  1851    1   2   0  97
  13   56   0   19   731  333 1023    1   54  643    1  1605    1   2   0  97
  14   80   0   20   729  337 1040    1   49  702    1  1801    1   2   0  97
  15   58   0   16   668  299  942    1   52  604    1  1520    1   2   0  97

[gopherbot]

Comment 5 by zheganin@enaza.ru:

Yeah, looks pretty much like it.
Can be mitigated by creating a processor set and building go on it, then running go
programs on it.

[gopherbot]

Comment 6 by zheganin@enaza.ru:

(Just to add an explicit feedback: it worked for me).

[4ad]

Comment 7:

Note that this failure only occurs in the runtime tests and we have never seen it occur
in practice. Also, if you only want to build Go without running the tests, you can run
make.bash instead of all.bash.

Owner changed to @4ad.

Status changed to Accepted.

[4ad]

Comment 8:

Rewriting the runtime in Go fixed this.

Status changed to Fixed.