I'm attempting to create a wrapper for MPI, however, MPI_Init
wants to read the arguments for main():

MPI_Init(int *argv, char ***argc);

How do I get this last level of pointer reference?

So far, I have:

void main (string[] args)
{
     auto argarr = new char*[args.length];
     foreach(i, a; args)
         argarr[i] = (a.dup ~ '\0').ptr;

     int argc = to!(int)(argarr.length);
     MPI_Init(&argc, argarr.ptr);
}

Any ideas?

-Andrew

On Sun, 12 Aug 2012 22:30:57 +0200, Andrew <andrew.spott@gmail.com> wrote:

> I'm attempting to create a wrapper for MPI, however, MPI_Init
> wants to read the arguments for main():
>
> MPI_Init(int *argv, char ***argc);
>
> How do I get this last level of pointer reference?
>
> So far, I have:
>
> void main (string[] args)
> {
>       auto argarr = new char*[args.length];
>       foreach(i, a; args)
>           argarr[i] = (a.dup ~ '\0').ptr;
>
>       int argc = to!(int)(argarr.length);
>       MPI_Init(&argc, argarr.ptr);
> }
>
> Any ideas?


    // Array of pointers to command line parameters.
    char*[] argv = args.map!((a)=>(a.dup~'\0').ptr).array;
    // Address of first element of that array.
    char** argvp = &argv[0];
    // And finally, address of the pointer to that first element.
    char*** argvpp = &argvp;

Now, the interesting part is reconstructing the string[] from
the potentially modified argvpp...

-- 
Simen

Would this help?
https://github.com/1100110/OpenMPI

On Sun, 12 Aug 2012 15:30:57 -0500, Andrew <andrew.spott@gmail.com> wrote:

> I'm attempting to create a wrapper for MPI, however, MPI_Init
> wants to read the arguments for main():
>
> MPI_Init(int *argv, char ***argc);
>
> How do I get this last level of pointer reference?
>
> So far, I have:
>
> void main (string[] args)
> {
>       auto argarr = new char*[args.length];
>       foreach(i, a; args)
>           argarr[i] = (a.dup ~ '\0').ptr;
>
>       int argc = to!(int)(argarr.length);
>       MPI_Init(&argc, argarr.ptr);
> }
>
> Any ideas?
>
> -Andrew


-- 
Using Opera's revolutionary email client: http://www.opera.com/mail/

Also the Elementary repo has a wrapper for main() from E17, might be useful.

On Sun, 12 Aug 2012 15:54:45 -0500, 1100110 <10equals2@gmail.com> wrote:

> Would this help?
> https://github.com/1100110/OpenMPI
>
> On Sun, 12 Aug 2012 15:30:57 -0500, Andrew <andrew.spott@gmail.com> wrote:
>
>> I'm attempting to create a wrapper for MPI, however, MPI_Init
>> wants to read the arguments for main():
>>
>> MPI_Init(int *argv, char ***argc);
>>
>> How do I get this last level of pointer reference?
>>
>> So far, I have:
>>
>> void main (string[] args)
>> {
>>       auto argarr = new char*[args.length];
>>       foreach(i, a; args)
>>           argarr[i] = (a.dup ~ '\0').ptr;
>>
>>       int argc = to!(int)(argarr.length);
>>       MPI_Init(&argc, argarr.ptr);
>> }
>>
>> Any ideas?
>>
>> -Andrew
>
>


-- 
Using Opera's revolutionary email client: http://www.opera.com/mail/

On 8/12/12, Simen Kjaeraas <simen.kjaras@gmail.com> wrote:
>      // Array of pointers to command line parameters.
>      char*[] argv = args.map!((a)=>(a.dup~'\0').ptr).array;

You know.. it'd be much simpler if argc & argv were stored somewhere. druntime/src/rt/dmain2.d is where the action begins:

extern (C) int main(int argc, char** argv) { ... }

And then this is stored as a string[] to _d_args which can be picked up via Runtime.args() in core.runtime.

I'm thinking that having to retrieve the original argc/argv is common when interfacing with C/C++, so maybe we should have a druntime function which can return the original unprocessed args?

On Sun, 12 Aug 2012 23:02:43 +0200, Andrej Mitrovic <andrej.mitrovich@gmail.com> wrote:

> On 8/12/12, Simen Kjaeraas <simen.kjaras@gmail.com> wrote:
>>      // Array of pointers to command line parameters.
>>      char*[] argv = args.map!((a)=>(a.dup~'\0').ptr).array;
>
> You know.. it'd be much simpler if argc & argv were stored somewhere.
> druntime/src/rt/dmain2.d is where the action begins:
>
> extern (C) int main(int argc, char** argv) { ... }
>
> And then this is stored as a string[] to _d_args which can be picked
> up via Runtime.args() in core.runtime.
>
> I'm thinking that having to retrieve the original argc/argv is common
> when interfacing with C/C++, so maybe we should have a druntime
> function which can return the original unprocessed args?

That don't sound too stupid. File an enhancement request, wouldya?

-- 
Simen

Would this be useful?
https://github.com/1100110/OpenMPI

On Sunday, 12 August 2012 at 20:52:03 UTC, Simen Kjaeraas wrote:
> On Sun, 12 Aug 2012 22:30:57 +0200, Andrew <andrew.spott@gmail.com> wrote:
>
>> I'm attempting to create a wrapper for MPI, however, MPI_Init
>> wants to read the arguments for main():
>>
>> MPI_Init(int *argv, char ***argc);
>>
>> How do I get this last level of pointer reference?
>>
>> So far, I have:
>>
>> void main (string[] args)
>> {
>>      auto argarr = new char*[args.length];
>>      foreach(i, a; args)
>>          argarr[i] = (a.dup ~ '\0').ptr;
>>
>>      int argc = to!(int)(argarr.length);
>>      MPI_Init(&argc, argarr.ptr);
>> }
>>
>> Any ideas?
>
>
>     // Array of pointers to command line parameters.
>     char*[] argv = args.map!((a)=>(a.dup~'\0').ptr).array;
>     // Address of first element of that array.
>     char** argvp = &argv[0];
>     // And finally, address of the pointer to that first element.
>     char*** argvpp = &argvp;
>
> Now, the interesting part is reconstructing the string[] from
> the potentially modified argvpp...

(sorry for spam...)

On Sunday, 12 August 2012 at 21:04:55 UTC, 1100110 wrote:
> Would this be useful?
> https://github.com/1100110/OpenMPI
>
> On Sunday, 12 August 2012 at 20:52:03 UTC, Simen Kjaeraas wrote:
>> On Sun, 12 Aug 2012 22:30:57 +0200, Andrew <andrew.spott@gmail.com> wrote:
>>
>>> I'm attempting to create a wrapper for MPI, however, MPI_Init
>>> wants to read the arguments for main():
>>>
>>> MPI_Init(int *argv, char ***argc);
>>>
>>> How do I get this last level of pointer reference?
>>>
>>> So far, I have:
>>>
>>> void main (string[] args)
>>> {
>>>     auto argarr = new char*[args.length];
>>>     foreach(i, a; args)
>>>         argarr[i] = (a.dup ~ '\0').ptr;
>>>
>>>     int argc = to!(int)(argarr.length);
>>>     MPI_Init(&argc, argarr.ptr);
>>> }
>>>
>>> Any ideas?
>>
>>
>>    // Array of pointers to command line parameters.
>>    char*[] argv = args.map!((a)=>(a.dup~'\0').ptr).array;
>>    // Address of first element of that array.
>>    char** argvp = &argv[0];
>>    // And finally, address of the pointer to that first element.
>>    char*** argvpp = &argvp;
>>
>> Now, the interesting part is reconstructing the string[] from
>> the potentially modified argvpp...

On 8/12/12, Simen Kjaeraas <simen.kjaras@gmail.com> wrote:
> That don't sound too stupid. File an enhancement request, wouldya?

http://d.puremagic.com/issues/show_bug.cgi?id=8544

Yes, incredibly.

Thanks!

On Sunday, 12 August 2012 at 20:57:47 UTC, 1100110 wrote:
> Would this help?
> https://github.com/1100110/OpenMPI
>
> On Sun, 12 Aug 2012 15:30:57 -0500, Andrew <andrew.spott@gmail.com> wrote:
>
>> I'm attempting to create a wrapper for MPI, however, MPI_Init
>> wants to read the arguments for main():
>>
>> MPI_Init(int *argv, char ***argc);
>>
>> How do I get this last level of pointer reference?
>>
>> So far, I have:
>>
>> void main (string[] args)
>> {
>>      auto argarr = new char*[args.length];
>>      foreach(i, a; args)
>>          argarr[i] = (a.dup ~ '\0').ptr;
>>
>>      int argc = to!(int)(argarr.length);
>>      MPI_Init(&argc, argarr.ptr);
>> }
>>
>> Any ideas?
>>
>> -Andrew