Hi,

I noticed that sometimes on windows this line of code crashes the test suites.

https://github.com/gedaiu/fluent-asserts/blob/master/core/fluentasserts/core/results.d#L1072

This exception can be captured only with a debugger... is it a dmd bug?

```
Thread 25CC created, Entry: mswsock.00007FF9011604C0
EXCEPTION_DEBUG_INFO:
           dwFirstChance: 1
           ExceptionCode: C0000005 (EXCEPTION_ACCESS_VIOLATION)
          ExceptionFlags: 00000000
        ExceptionAddress: 00007FF9055E302C ntdll.00007FF9055E302C
        NumberParameters: 2
ExceptionInformation[00]: 0000000000000000 Read
ExceptionInformation[01]: FFFFFFFFFFFFFFFF Inaccessible Address
First chance exception on 00007FF9055E302C (C0000005, EXCEPTION_ACCESS_VIOLATION)!
```

I think this is related with this issue:
https://github.com/gedaiu/fluent-asserts/issues/63

Can anyone help me with this? I don't even know how to start fixing this issue...

Thanks,
Bogdan

writeln(fileName);
if (!fileName.exists)
{
  return;
}

:)

On Thursday, 14 December 2017 at 14:47:25 UTC, Kagamin wrote:
> writeln(fileName);
> if (!fileName.exists)
> {
>   return;
> }
>
> :)

I'm not sure I understand this solution...

Try printf debugging in case argument is invalid.

On Friday, 15 December 2017 at 09:24:07 UTC, Kagamin wrote:
> Try printf debugging in case argument is invalid.

ah .. ok ...

I tried to debug the issue and it looks like the filename is valid and there is no null value.

I am thinking that the value is destroyed before it reach the `GetFileAttributesW` or during that call... but I don't see how that is possible.

You can also try to call `exists` somewhere before this part of code.

You said tests fail?

class SourceResult
{
	private const
	{
		string file;
		size_t line;
	}
	this(string fileName = __FILE__, size_t line = __LINE__, size_t range = 6) nothrow
	{
		this.file = fileName;
		this.line = line;
		if (!fileName.exists)
		{
			return;
		}
	}
}

unittest
{
	auto result = new SourceResult("test/values.d", 26);
	auto msg = result.file;
}

Does this fail too?

That said, tempCString code is suspicious:
https://github.com/dlang/phobos/blob/master/std/internal/cstring.d#L221
If unittest-versioned exists calls release-versioned tempCString, it will corrupt the stack. Try to replace 16 with 256 there and recompile your code.

Maybe this https://issues.dlang.org/show_bug.cgi?id=18084

On 12/15/17 10:08 AM, Kagamin wrote:
> Maybe this https://issues.dlang.org/show_bug.cgi?id=18084

Thanks for looking into this. I created a PR to fix.

Szabo, can you please try with this patch and see if it fixes your issue?

https://github.com/dlang/phobos/pull/5932

-Steve