[kaffe] solaris status

[Dalibor Topic]

Ciao Riccardo,

Riccardo wrote:
> Hey,
> 
> as Dalibor suggested, I tried building on solaris with the interpreter 
> engine instead if JIT. To do this I updated to today's CVS (from oct 
> stuff), reconfigured as he suggested and built. Build stops with a 
> linking error as my page reports.
> 
> 
> http://homepage.mac.com/riccardo_mottola/kaffe-devel/machines/legolas/
> Kaffe_on_Solaris_2_6_sparc.html

The linking errors are certainly weird if you're building with the 
interpreter engine. I've tried grep -r on some of the symbols the linker 
is complaining about, but all symbols I tried were from 
kaffe/kaffevm/jit, and make shouldn't be using files from there when 
it's building an interpreter engine.

Could you do a anither build with the same configure options in a clean 
directory, and send me the full build log and config.log? I'd like to 
attempt to figure out what is being compiled and why ;)

cheers,
dalibor topic

p.s. my results with sparc-solaris 2.9, kaffe 1.1.3, jit engine:

================================
5 of 144 tests failed
Please report to kaffe at kaffe.org
================================

FAIL: DoubleCvt.java
FAIL: ThreadState.java
FAIL: NetworkInterfaceTest.java
FAIL: CLTestConc.java
FAIL: TestUnlock.java

In detail (the jit specific failures, the NetworkInterfaceTest fails on 
intrp as well):

/tmp/topic/build-jit [521] diff -u test/regression/DoubleCvt.[of]*
--- test/regression/DoubleCvt.fail      2003-12-09 18:48:22.648165250 +0100
+++ test/regression/DoubleCvt.out       2003-12-09 18:48:21.796667770 +0100
@@ -1,6 +1,6 @@
  NaN: 7ff8000000000000
  NaN as float: 7fc00000
-NaN as int: 2147483647 7fffffff
+NaN as int: 0 0
  NaN as long: 0 0
  MIN_VALUE: 1
  MIN_VALUE as float: 0
@@ -532,7 +532,7 @@
  lnp(64) as long: -1 -1
  NaN: 7fc00000
  NaN as double: 7ff8000000000000
-NaN as int: 2147483647 7fffffff
+NaN as int: 0 0
  NaN as long: 0 0
  MIN_VALUE: 1
  MIN_VALUE as double: 36a0000000000000

/tmp/topic/build-jit [522] diff -u test/regression/ThreadState.[of]*
--- test/regression/ThreadState.fail    2003-12-09 18:50:25.628003979 +0100
+++ test/regression/ThreadState.out     2003-12-09 18:49:24.597253608 +0100
@@ -1 +1,10 @@
-Time out.  Failure.
+Success.
+Success.
+Success.
+Success.
+Success.
+Success.
+Success.
+Success.
+Success.
+Success.

/tmp/topic/build-jit [523] diff -u test/regression/CLTestConc.[of]*
--- test/regression/CLTestConc.fail     2003-12-09 18:59:48.177672662 +0100
+++ test/regression/CLTestConc.out      2003-12-09 18:54:47.965283168 +0100
@@ -1 +1 @@
-Timeout - 300 seconds - killed
+Success.

/tmp/topic/build-jit [524] diff -u test/regression/TestUnlock.[of]*
--- test/regression/TestUnlock.fail     2003-12-09 19:01:15.179812905 +0100
+++ test/regression/TestUnlock.out      2003-12-09 19:01:04.800835603 +0100
@@ -1 +1 @@
-Time out.  Failure.
+Success.

There is clearly some problem with locking/unlocking that doesn't show 
up with the interpreter. Maybe the interpreter is just to slow for it to 
show up, though. The NaN bug looks plain weird to me, as it's not 
showing up in the interpreter. I'd be grateful for any clues into how to 
debug those failures.

cheers,
dalibor topic