https://issues.dlang.org/show_bug.cgi?id=18493

Radu <radu.racariu@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |betterC

--

https://issues.dlang.org/show_bug.cgi?id=18493

Radu <radu.racariu@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|enhancement                 |blocker

--- Comment #1 from Radu <radu.racariu@gmail.com> ---
Errors out on dmd 2.0.78.3 and dmd 2.0.79-beta1

https://run.dlang.io/is/R6BWSz

--

https://issues.dlang.org/show_bug.cgi?id=18493

--- Comment #2 from Radu Racariu <radu.racariu@gmail.com> ---
Even more reduced test case:
---
struct S
{
    this(this)
    {
    }

    ~this()
    {
    }
}

struct C
{
    S s1;
    S s2;
}
---

--

https://issues.dlang.org/show_bug.cgi?id=18493

Radu Racariu <radu.racariu@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |industry

--

https://issues.dlang.org/show_bug.cgi?id=18493

--- Comment #3 from Radu Racariu <radu.racariu@gmail.com> ---
Declaring `C` as

---
struct C
{
    S s1;
}
---

makes the test case compile.

--

https://issues.dlang.org/show_bug.cgi?id=18493

Mike Franklin <slavo5150@yahoo.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |pull
                 CC|                            |slavo5150@yahoo.com

--- Comment #4 from Mike Franklin <slavo5150@yahoo.com> ---
A PR has been submitted to remove generation of the try-catch if the postblit is `nothrow`:  https://github.com/dlang/dmd/pull/8184

I don't consider this a complete fix to this issue, but if it is accepted, it will allow one to attribute their postblit with `nothrow` to avoid the error in -betterC.

I'm pondering a more thorough solution for -betterC, and intend to tackle that as another PR.

--

https://issues.dlang.org/show_bug.cgi?id=18493

--- Comment #5 from Radu Racariu <radu.racariu@gmail.com> ---
Ideally -betterC should imply `nothrow`, and any lowered code should work with this assumption.

What is important to keep in mind is that if you require `this(this)` to be annotated with `nothrow` in order to compile - you need to fix the error message for the case where it isn't and using -betterC, the current error message is very bad.

--

https://issues.dlang.org/show_bug.cgi?id=18493

--- Comment #6 from Radu Racariu <radu.racariu@gmail.com> ---
I though more about this and I think betterC should imply nothrow *only for* compiler generated code.

Making betterC nothrow by default will change semantics and basically split the language in two, as betterC code will have 2 different defaults depending on the switch.

This means that betterC should start error out on non nothrow code, with a deprecation period for interim.

--

https://issues.dlang.org/show_bug.cgi?id=18493

--- Comment #7 from github-bugzilla@puremagic.com ---
Commit pushed to master at https://github.com/dlang/dmd

https://github.com/dlang/dmd/commit/8a33fb6c692d0b2c1f03ceeb65c28beaebd932a0 Add test for issue 18493

--

https://issues.dlang.org/show_bug.cgi?id=18493

Mike Franklin <slavo5150@yahoo.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://issues.dlang.org/sh
                   |                            |ow_bug.cgi?id=18828

--