Fixing this bug will be important for 64 bits:
3150 cast from dynamic array to ulong is allowed

As discussed in the comments to that bug, a change is required at two places in druntime, because druntime currently relies on the compiler bug. I don't think it's particularly difficult, but I don't want to do it myself. Can one of you guys take care of it?

I've sort of unofficially taken ownership of that module ;)

I'll try and fix it, it shouldn't be too difficult, there are some functions already that use a local Array struct, I'll just duplicate those efforts.

-Steve



----- Original Message ----
> From: Don Clugston <dclugston at googlemail.com>
> To: D-runtime at puremagic.com
> Sent: Fri, July 23, 2010 5:31:04 AM
> Subject: [D-runtime] Bug 3150: requires changes to rt/lifetime.d
> 
> Fixing this bug will be important for 64 bits:
> 3150 cast from dynamic array  to ulong is allowed
> 
> As discussed in the comments to that bug, a change is  required at two
> places in druntime, because druntime currently relies on the  compiler
> bug. I don't think it's particularly difficult, but I don't want to  do
> it myself. Can one of you guys take care of  it?
> _______________________________________________
> D-runtime mailing  list
> D-runtime at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/d-runtime
> 

If it helps, GDC has always worked the way we want.

Sent from my iPhone

On Jul 23, 2010, at 5:21 AM, Steve Schveighoffer <schveiguy at yahoo.com> wrote:

> I've sort of unofficially taken ownership of that module ;)
> 
> I'll try and fix it, it shouldn't be too difficult, there are some functions already that use a local Array struct, I'll just duplicate those efforts.
> 
> -Steve
> 
> 
> 
> ----- Original Message ----
>> From: Don Clugston <dclugston at googlemail.com>
>> To: D-runtime at puremagic.com
>> Sent: Fri, July 23, 2010 5:31:04 AM
>> Subject: [D-runtime] Bug 3150: requires changes to rt/lifetime.d
>> 
>> Fixing this bug will be important for 64 bits:
>> 3150 cast from dynamic array  to ulong is allowed
>> 
>> As discussed in the comments to that bug, a change is  required at two
>> places in druntime, because druntime currently relies on the  compiler
>> bug. I don't think it's particularly difficult, but I don't want to  do
>> it myself. Can one of you guys take care of  it?
>> _______________________________________________
>> D-runtime mailing  list
>> D-runtime at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/d-runtime
>> 
> 
> 
> 
> _______________________________________________
> D-runtime mailing list
> D-runtime at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/d-runtime

I have a problem here.  If I change a function that returns a ulong to having it return an Array struct, the compiler expects the caller to push a reference to the return value on the stack, where it doesn't do that when returning a ulong.

Easily demonstrated with this code:

testlongtoarray2.d:
struct Array
{
    size_t length;
    ubyte* ptr;
}

extern(C) Array callme()
{
    Array d;
    d.length = 5;
    d.ptr = null;
    return d;
}

testlongtoarray.d:
extern(C) ulong callme();

void main()
{
    ulong x = callme();
}
[steves at steveslaptop testd]$ dmd -gc testlongtoarray*.d[steves at steveslaptop
testd]$ gdb testlongtoarray2
GNU gdb (GDB) Fedora (6.8.50.20090302-40.fc11)
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "i586-redhat-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
(gdb) run
Starting program: /home/steves/testd/testlongtoarray2
[Thread debugging using libthread_db enabled]

Program received signal SIGSEGV, Segmentation fault.
0x08049073 in callme (__HID1=0x180fc4) at testlongtoarray2.d:9  <------ note the
hidden argument which is garbage!
9        Array d;
Missing separate debuginfos, use: debuginfo-install glibc-2.10.2-1.i686
(gdb)

What's happening is that main is simply calling the function without pushing anything on the stack, but the implementation expects that the reference to the return value is on the stack.

Does this make any sense to anyone?  What should I do here?  I can't really change what the compiler expects the return value to be.

-Steve



----- Original Message ----
> From: Don Clugston <dclugston at googlemail.com>
> To: D-runtime at puremagic.com
> Sent: Fri, July 23, 2010 5:31:04 AM
> Subject: [D-runtime] Bug 3150: requires changes to rt/lifetime.d
> 
> Fixing this bug will be important for 64 bits:
> 3150 cast from dynamic array  to ulong is allowed
> 
> As discussed in the comments to that bug, a change is  required at two
> places in druntime, because druntime currently relies on the  compiler
> bug. I don't think it's particularly difficult, but I don't want to  do
> it myself. Can one of you guys take care of  it?
> _______________________________________________
> D-runtime mailing  list
> D-runtime at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/d-runtime
> 

That's why I never changed it myself :-)  The compiler needs to be changed to use the new functions.

On Jul 23, 2010, at 8:54 AM, Steve Schveighoffer wrote:

> I have a problem here.  If I change a function that returns a ulong to having it return an Array struct, the compiler expects the caller to push a reference to the return value on the stack, where it doesn't do that when returning a ulong.
> 
> Easily demonstrated with this code:
> 
> testlongtoarray2.d:
> struct Array
> {
>    size_t length;
>    ubyte* ptr;
> }
> 
> extern(C) Array callme()
> {
>    Array d;
>    d.length = 5;
>    d.ptr = null;
>    return d;
> }
> 
> testlongtoarray.d:
> extern(C) ulong callme();
> 
> void main()
> {
>    ulong x = callme();
> }
> [steves at steveslaptop testd]$ dmd -gc testlongtoarray*.d[steves at steveslaptop
> testd]$ gdb testlongtoarray2
> GNU gdb (GDB) Fedora (6.8.50.20090302-40.fc11)
> Copyright (C) 2009 Free Software Foundation, Inc.
> License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
> This is free software: you are free to change and redistribute it.
> There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
> and "show warranty" for details.
> This GDB was configured as "i586-redhat-linux-gnu".
> For bug reporting instructions, please see:
> <http://www.gnu.org/software/gdb/bugs/>...
> (gdb) run
> Starting program: /home/steves/testd/testlongtoarray2
> [Thread debugging using libthread_db enabled]
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x08049073 in callme (__HID1=0x180fc4) at testlongtoarray2.d:9  <------ note the
> hidden argument which is garbage!
> 9        Array d;
> Missing separate debuginfos, use: debuginfo-install glibc-2.10.2-1.i686
> (gdb)
> 
> What's happening is that main is simply calling the function without pushing anything on the stack, but the implementation expects that the reference to the return value is on the stack.
> 
> Does this make any sense to anyone?  What should I do here?  I can't really change what the compiler expects the return value to be.
> 
> -Steve
> 
> 
> 
> ----- Original Message ----
>> From: Don Clugston <dclugston at googlemail.com>
>> To: D-runtime at puremagic.com
>> Sent: Fri, July 23, 2010 5:31:04 AM
>> Subject: [D-runtime] Bug 3150: requires changes to rt/lifetime.d
>> 
>> Fixing this bug will be important for 64 bits:
>> 3150 cast from dynamic array  to ulong is allowed
>> 
>> As discussed in the comments to that bug, a change is  required at two
>> places in druntime, because druntime currently relies on the  compiler
>> bug. I don't think it's particularly difficult, but I don't want to  do
>> it myself. Can one of you guys take care of  it?
>> _______________________________________________
>> D-runtime mailing  list
>> D-runtime at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/d-runtime
>> 
> 
> 
> 
> _______________________________________________
> D-runtime mailing list
> D-runtime at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/d-runtime

You can still make the change I suggested in the bug report, adding a
static assert to check that the sizes are the same, and doing a cast
to ulong via a cast(void *), instead of a direct cast.
BTW I can't see any reason for any of those functions to use C calling
conventions. I presume they only reason they use C is because they're
so ancient.


On 23 July 2010 18:47, Sean Kelly <sean at invisibleduck.org> wrote:
> That's why I never changed it myself :-) ?The compiler needs to be changed to use the new functions.
>
> On Jul 23, 2010, at 8:54 AM, Steve Schveighoffer wrote:
>
>> I have a problem here. ?If I change a function that returns a ulong to having it return an Array struct, the compiler expects the caller to push a reference to the return value on the stack, where it doesn't do that when returning a ulong.
>>
>> Easily demonstrated with this code:
>>
>> testlongtoarray2.d:
>> struct Array
>> {
>> ? ?size_t length;
>> ? ?ubyte* ptr;
>> }
>>
>> extern(C) Array callme()
>> {
>> ? ?Array d;
>> ? ?d.length = 5;
>> ? ?d.ptr = null;
>> ? ?return d;
>> }
>>
>> testlongtoarray.d:
>> extern(C) ulong callme();
>>
>> void main()
>> {
>> ? ?ulong x = callme();
>> }
>> [steves at steveslaptop testd]$ dmd -gc testlongtoarray*.d[steves at steveslaptop
>> testd]$ gdb testlongtoarray2
>> GNU gdb (GDB) Fedora (6.8.50.20090302-40.fc11)
>> Copyright (C) 2009 Free Software Foundation, Inc.
>> License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
>> This is free software: you are free to change and redistribute it.
>> There is NO WARRANTY, to the extent permitted by law. ?Type "show copying"
>> and "show warranty" for details.
>> This GDB was configured as "i586-redhat-linux-gnu".
>> For bug reporting instructions, please see:
>> <http://www.gnu.org/software/gdb/bugs/>...
>> (gdb) run
>> Starting program: /home/steves/testd/testlongtoarray2
>> [Thread debugging using libthread_db enabled]
>>
>> Program received signal SIGSEGV, Segmentation fault.
>> 0x08049073 in callme (__HID1=0x180fc4) at testlongtoarray2.d:9 ?<------ note the
>> hidden argument which is garbage!
>> 9 ? ? ? ?Array d;
>> Missing separate debuginfos, use: debuginfo-install glibc-2.10.2-1.i686
>> (gdb)
>>
>> What's happening is that main is simply calling the function without pushing anything on the stack, but the implementation expects that the reference to the return value is on the stack.
>>
>> Does this make any sense to anyone? ?What should I do here? ?I can't really change what the compiler expects the return value to be.
>>
>> -Steve
>>
>>
>>
>> ----- Original Message ----
>>> From: Don Clugston <dclugston at googlemail.com>
>>> To: D-runtime at puremagic.com
>>> Sent: Fri, July 23, 2010 5:31:04 AM
>>> Subject: [D-runtime] Bug 3150: requires changes to rt/lifetime.d
>>>
>>> Fixing this bug will be important for 64 bits:
>>> 3150 cast from dynamic array ?to ulong is allowed
>>>
>>> As discussed in the comments to that bug, a change is ?required at two
>>> places in druntime, because druntime currently relies on the ?compiler
>>> bug. I don't think it's particularly difficult, but I don't want to ?do
>>> it myself. Can one of you guys take care of ?it?
>>> _______________________________________________
>>> D-runtime mailing ?list
>>> D-runtime at puremagic.com
>>> http://lists.puremagic.com/mailman/listinfo/d-runtime
>>>
>>
>>
>>
>> _______________________________________________
>> D-runtime mailing list
>> D-runtime at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/d-runtime
>
> _______________________________________________
> D-runtime mailing list
> D-runtime at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/d-runtime
>

It does allow them to be moved to a different module without compiler changes, but otherwise you're right.

On Jul 24, 2010, at 1:44 AM, Don Clugston wrote:

> You can still make the change I suggested in the bug report, adding a
> static assert to check that the sizes are the same, and doing a cast
> to ulong via a cast(void *), instead of a direct cast.
> BTW I can't see any reason for any of those functions to use C calling
> conventions. I presume they only reason they use C is because they're
> so ancient.
> 
> 
> On 23 July 2010 18:47, Sean Kelly <sean at invisibleduck.org> wrote:
>> That's why I never changed it myself :-)  The compiler needs to be changed to use the new functions.
>> 
>> On Jul 23, 2010, at 8:54 AM, Steve Schveighoffer wrote:
>> 
>>> I have a problem here.  If I change a function that returns a ulong to having it return an Array struct, the compiler expects the caller to push a reference to the return value on the stack, where it doesn't do that when returning a ulong.
>>> 
>>> Easily demonstrated with this code:
>>> 
>>> testlongtoarray2.d:
>>> struct Array
>>> {
>>>    size_t length;
>>>    ubyte* ptr;
>>> }
>>> 
>>> extern(C) Array callme()
>>> {
>>>    Array d;
>>>    d.length = 5;
>>>    d.ptr = null;
>>>    return d;
>>> }
>>> 
>>> testlongtoarray.d:
>>> extern(C) ulong callme();
>>> 
>>> void main()
>>> {
>>>    ulong x = callme();
>>> }
>>> [steves at steveslaptop testd]$ dmd -gc testlongtoarray*.d[steves at steveslaptop
>>> testd]$ gdb testlongtoarray2
>>> GNU gdb (GDB) Fedora (6.8.50.20090302-40.fc11)
>>> Copyright (C) 2009 Free Software Foundation, Inc.
>>> License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
>>> This is free software: you are free to change and redistribute it.
>>> There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
>>> and "show warranty" for details.
>>> This GDB was configured as "i586-redhat-linux-gnu".
>>> For bug reporting instructions, please see:
>>> <http://www.gnu.org/software/gdb/bugs/>...
>>> (gdb) run
>>> Starting program: /home/steves/testd/testlongtoarray2
>>> [Thread debugging using libthread_db enabled]
>>> 
>>> Program received signal SIGSEGV, Segmentation fault.
>>> 0x08049073 in callme (__HID1=0x180fc4) at testlongtoarray2.d:9  <------ note the
>>> hidden argument which is garbage!
>>> 9        Array d;
>>> Missing separate debuginfos, use: debuginfo-install glibc-2.10.2-1.i686
>>> (gdb)
>>> 
>>> What's happening is that main is simply calling the function without pushing anything on the stack, but the implementation expects that the reference to the return value is on the stack.
>>> 
>>> Does this make any sense to anyone?  What should I do here?  I can't really change what the compiler expects the return value to be.
>>> 
>>> -Steve
>>> 
>>> 
>>> 
>>> ----- Original Message ----
>>>> From: Don Clugston <dclugston at googlemail.com>
>>>> To: D-runtime at puremagic.com
>>>> Sent: Fri, July 23, 2010 5:31:04 AM
>>>> Subject: [D-runtime] Bug 3150: requires changes to rt/lifetime.d
>>>> 
>>>> Fixing this bug will be important for 64 bits:
>>>> 3150 cast from dynamic array  to ulong is allowed
>>>> 
>>>> As discussed in the comments to that bug, a change is  required at two
>>>> places in druntime, because druntime currently relies on the  compiler
>>>> bug. I don't think it's particularly difficult, but I don't want to  do
>>>> it myself. Can one of you guys take care of  it?
>>>> _______________________________________________
>>>> D-runtime mailing  list
>>>> D-runtime at puremagic.com
>>>> http://lists.puremagic.com/mailman/listinfo/d-runtime
>>>> 
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> D-runtime mailing list
>>> D-runtime at puremagic.com
>>> http://lists.puremagic.com/mailman/listinfo/d-runtime
>> 
>> _______________________________________________
>> D-runtime mailing list
>> D-runtime at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/d-runtime
>> 
> _______________________________________________
> D-runtime mailing list
> D-runtime at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/d-runtime