Christian Thalinger wrote:
On Fri, 2006-01-06 at 02:33 +0100, Dalibor Topic wrote:
Hi all,
I've cleaned up a bit in fdlibm, removing the use of gcj-specific
java-assert.h headr and replaced it by plain asserts. The former code
caused problems on platforms like Cygwin when built in debug mode, as it
tried to link to a method that does not exist in GNU Classpath.
cheers,
dalibor topic
2006-01-06 Dalibor Topic <[EMAIL PROTECTED]>
* native/fdlibm/Makefile.am (libfdlibm_la_SOURCES): Removed
java-assert.h.
* native/fdlibm/java-assert.h: Removed file.
* native/fdlibm/mprec.c: Include assert.h. Don't include
java-assert.h.
Replaced use of JvAssert by assert.
Hmm, i'm not sure if actually this change causes the problem, but the
CACAO fptest does not work anymore:
925. 5.0 / MAX = double: cacao: mprec.c:100: _Jv_Balloc: Assertion
`(1 << k) < 128' failed.
Program received signal SIGABRT, Aborted.
[Switching to Thread 46912504037360 (LWP 27552)]
0x00002aaaab028df0 in raise () from /lib/libc.so.6
(gdb) bt
#0 0x00002aaaab028df0 in raise () from /lib/libc.so.6
#1 0x00002aaaab02a2a0 in abort () from /lib/libc.so.6
#2 0x00002aaaab021ffb in __assert_fail () from /lib/libc.so.6
#3 0x00002aaaabbb4e59 in _Jv_Balloc (ptr=Variable "ptr" is not
available.
) at mprec.c:100
Maybe the old JvAssert did not actually do anything?
JvAssert only kicked in if DEBUG was defined, which assert always kicks
in unless NDEBUG is defined. Other than that, they do the same checks,
so it may be a real bug in mprec.c. Could you investigate on your arch/os?
cheers,
dalibor topic
_______________________________________________
Classpath-patches mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/classpath-patches
