string fileContent = "";

...

writeln(ri, ": debug 1");
foreach(i; 0..dim)
{
if( i > 0 ){ fileContent ~= "\t"; }

writeln(ri, ": debug 1.1: ", ri*dim + i, ": ", positions[ ri*dim + i ]);

fileContent ~= to!string(positions[ ri*dim + i ]);

writeln(ri, ": debug 1.2: ", ri*dim + i, ": ", positions[ ri*dim + i ]);

}

On line "fileContent ~= ...", I get a segmentation fault.

"positions" array is defined as auto positions = new float[ 100 ]; So, I am 100% sure, it is not out of range. "ri*dim + 1" is not a big number at all.

...
4: debug 1.1: 9: 0.271075
4: debug 1.2: 9: 0.271075
4: debug 2
4: debug 2.1: 4
4: debug 3
4: debug 4
5: debug 1
5: debug 1.1: 10: 0.884978
5: debug 1.2: 10: 0.884978
5: debug 1.1: 11: 0.813104
Segmentation fault
...

I have compiled the code with "-g" flag and ran it with GNU debugger. It gives following:

Thread 1 "dataspace" received signal SIGSEGV, Segmentation fault.
0x00005555556ca286 in _D2gc4impl12conservativeQw3Gcx10smallAllocMFNbmKmkxC8TypeInfoZPv ()

So, there is a problem about small allocation. I remember I had this problem before in another project.

I have enough free ram. htop shows 3.96 GiB of 8 GiB is used only and swap is not in use.

DMD64 D Compiler v2.094.0

Is this error related to me? Is it a programmer error? Is it a bug? Am I doing something wrong? This is a compiler related operation (string concatenation), and I assume/expect that it would work without a problem.

In other parts of the code, concatenation operations are all failing with same error. I need guidance to get out of this situation. My assumption was that as long as there is empty heap memory, concatenation operation would succeed always. But, it doesn't seem like so.

Are there any other threads in your program?

On Thursday, 22 April 2021 at 21:15:48 UTC, tcak wrote:

Oh and where is that positions variable defined?

On Friday, 23 April 2021 at 00:30:02 UTC, Adam D. Ruppe wrote:

I am doing OpenCL programming. CPU side is single threaded.

As far as I see, it is not related to that array or indices at all. After running for a short time, if a piece of code does any string/char or byte array concatenation at all, all of them cause segmentation fault with error:

_D2gc4impl12conservativeQw3Gcx10smallAllocMFNbmKmkxC8TypeInfoZPv ()

When I comment out those piece of codes, there is no error.

If there is no known situation that would cause this, I will need to update codes to C-style pre-allocate buffer and copy inside it instead of concatenating data.

On Friday, 23 April 2021 at 00:44:58 UTC, tcak wrote:

The question of where is to see if it was CTFE allocated or runtime allocated. I don't think it should make a difference here but idk.

druntime not being initialized is the only situation I know of that causes this directly. Otherwise memory corruption or race condition of some sort, hence why i was wondering about threads.

But I don't think it is uninitialized druntime unless that new happened at ctfe or something.

On Thursday, 22 April 2021 at 21:15:48 UTC, tcak wrote:

Do you have a minimal reproducible test case? 🤔