phobos commit, revision 2116


user: Don Clugston

msg:
Fix for BigInt assign for 64 bit compilers. Also a few steps towards const correctness, and some comment fixes.

http://www.dsource.org/projects/phobos/changeset/2116

dsource.org wrote:
> phobos commit, revision 2116
>
>
> user: Don Clugston
>
> msg:
> Fix for BigInt assign for 64 bit compilers. Also a few steps towards const correctness, and some comment fixes.
>
> http://www.dsource.org/projects/phobos/changeset/2116
>
> 

Still dies (OSX only):


core.exception.AssertError at std.bigint(481): unittest failure
----------------
5   bigint                              0x0000a918 onAssertErrorMsg + 68
6   bigint                              0x0000a95a onUnittestErrorMsg + 26
7   bigint                              0x00011f99 _d_unittestm + 45
8   bigint                              0x00001f03 void
std.bigint.__unittest_fa
il(int) + 35
9   bigint                              0x00002714 void
std.bigint.__unittest1()
 + 744
10  bigint                              0x00001e8b void
std.bigint.__modtest() +
 11
11  bigint                              0x0000ad25 extern (C) bool
core.runtime.
runModuleUnitTests() + 45
12  bigint                              0x00005b47
D6object10ModuleInfo7opApplyF
MDFKPS6object10ModuleInfoZiZi + 71
13  bigint                              0x0000ac16 runModuleUnitTests + 134
14  bigint                              0x00012302 extern (C) int
rt.dmain2.main
(int, char**) + 38
15  bigint                              0x00012206 extern (C) int
rt.dmain2.main
(int, char**) + 42
16  bigint                              0x00012193 main + 179
17  bigint                              0x00001e6d start + 53
make[2]: *** [generated/osx/debug/unittest/std/bigint] Error 1
make[1]: *** [unittest] Error 2
make: *** [unittest] Error 2

On 25 October 2010 09:45, Walter Bright <walter at digitalmars.com> wrote:
>
>
> dsource.org wrote:
>>
>> phobos commit, revision 2116
>>
>>
>> user: Don Clugston
>>
>> msg:
>> Fix for BigInt assign for 64 bit compilers. Also a few steps towards const
>> correctness, and some comment fixes.
>>
>> http://www.dsource.org/projects/phobos/changeset/2116
>>
>>
>
> Still dies (OSX only):
>
> core.exception.AssertError at std.bigint(481): unittest failure
>

Please add this line at 481, and tell me what it prints.

    assert(BigInt(0x1234_5678_9ABC_5A5AL).ulongLength == 1);
+    printf("%Lx%Lx\n", BigInt(0x1234_5678_9ABC_5A5AL).data.peekUlong(0));
    assert(BigInt(0x1234_5678_9ABC_5A5AL).toLong() == 0x1234_5678_9ABC_5A5AL);

obviously adding
import core.stdc.stdio;

It's looking to me as though have a long-related codegen bug.

An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20101025/25fc30d7/attachment.html>

Cool. This is definitely a compiler bug; it's negating the long before it stores it. Maybe this is a reduced test case:

void foo(T : long)(T x)
{
	ulong q = cast(ulong)((x < 0) ? -x : x);
	assert(q == 0x1234_5678_9ABC_5A5AL);
}

void bar(T : long)(T x)
{
	foo(x);
}

void main()
{
   bar(0x1234_5678_9ABC_5A5AL);
}
If that passes, it may be a problem with opAssign or with passing
ulong as a function parameter between modules.

That passes on OSX.

Don Clugston wrote:
> Cool. This is definitely a compiler bug; it's negating the long before it stores it. Maybe this is a reduced test case:
>
> void foo(T : long)(T x)
> {
> 	ulong q = cast(ulong)((x < 0) ? -x : x);
> 	assert(q == 0x1234_5678_9ABC_5A5AL);
> }
>
> void bar(T : long)(T x)
> {
> 	foo(x);
> }
>
> void main()
> {
>    bar(0x1234_5678_9ABC_5A5AL);
> }
> If that passes, it may be a problem with opAssign or with passing
> ulong as a function parameter between modules.
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
>
>
>