Bug 10223 – Variant[] within Variant[] can be accessed, but not assigned to

I've spent a lot of time trying to get workarounds that don't suck completely, but in the end I still expect this should "Just Work". import std.stdio; import std.variant; void main(){ Variant[] top, bottom; top = new Variant[](1); bottom = new Variant[](1); bottom[0] = "bar"; top[0] = bottom; writeln(top[0][0]); // This works top[0][0] = "foo"; // This does not writeln(top[0][0]); } //Output: bar std.variant.VariantException@std/variant.d(1231): Variant: attempting to use incompatible types immutable(char)[] and std.variant.VariantN!(32LU).VariantN ---------------- ./variantNest(@trusted long std.variant.VariantN!(32uL).VariantN.handler!(std.variant.VariantN!(32uL).VariantN[]).handler(std.variant.VariantN!(32uL).VariantN.OpID, ubyte[32]*, void*)+0x4e7) [0x44324f] ./variantNest(@trusted std.variant.VariantN!(32uL).VariantN std.variant.VariantN!(32uL).VariantN.opIndexAssign!(immutable(char)[], int).opIndexAssign(immutable(char)[], int)+0x10f) [0x447757] ./variantNest(_Dmain+0x245) [0x43b40d] ./variantNest(extern (C) int rt.dmain2._d_run_main(int, char**, extern (C) int function(char[][])*).void runMain()+0x18) [0x44d884] ./variantNest(extern (C) int rt.dmain2._d_run_main(int, char**, extern (C) int function(char[][])*).void tryExec(scope void delegate())+0x2a) [0x44d3b6] ./variantNest(extern (C) int rt.dmain2._d_run_main(int, char**, extern (C) int function(char[][])*).void runAll()+0x40) [0x44d8d4] ./variantNest(extern (C) int rt.dmain2._d_run_main(int, char**, extern (C) int function(char[][])*).void tryExec(scope void delegate())+0x2a) [0x44d3b6] ./variantNest(_d_run_main+0x1ae) [0x44d372] ./variantNest(main+0x17) [0x44d1bf] /lib64/libc.so.6(__libc_start_main+0xfd) [0x7fad4d2a44bd] ----------------

An NG post reminded me of this, so I just tested it on 2.067.1 and the output is rather different: wyatt@Yue ~/tmp/d $ rdmd ./issue10223.d bar Segmentation fault wyatt@Yue ~/tmp/d $ Okay, so it compiles now, but still doesn't work. We don't get nice things on Linux, so I tried it on my Windows workstation over lunch and found it's just been turned from a disappointment at compile time to a knife in the back at runtime: [0s] jwe29@PC91534 ~/tmp $ rdmd ./issue10223.d object.Error@(0): Access Violation ---------------- 0x0042B626 0x00405141 0x00404667 0x004073D2 0x004021B1 0x0040B5E6 0x0040B5BB 0x0040B4D3 0x0040750B 0x7480337A in BaseThreadInitThunk 0x76F28FE2 in RtlInitializeExceptionChain 0x76F28FB5 in RtlInitializeExceptionChain === Bypassed === std.variant.VariantException@std\variant.d(1445): Variant: attempting to use incompatible types immutable(char)[] and std.variant.VariantN!20u.VariantN ---------------- 0x004050E6 0x00404667 0x004073D2 0x004021B1 0x0040B5E6 0x0040B5BB 0x0040B4D3 0x0040750B 0x7480337A in BaseThreadInitThunk 0x76F28FE2 in RtlInitializeExceptionChain 0x76F28FB5 in RtlInitializeExceptionChain object.Error@(0): Access Violation ---------------- 0x0042B626 0x00405141 0x00404667 0x004073D2 0x004021B1 0x0040B5E6 0x0040B5BB 0x0040B4D3 0x0040750B 0x7480337A in BaseThreadInitThunk 0x76F28FE2 in RtlInitializeExceptionChain 0x76F28FB5 in RtlInitializeExceptionChain === ~Bypassed === bar [2s] jwe29@PC91534 ~/tmp $

I tried to hunt the bug down, but did only succeed partially. I'll write here, what I could find out. Maybe this is of some help for someone else: * The crash happens inside of the @trusted part of the destructor of VariantN. * When top[0][0] = "foo" is executed, opAssign is called twice, first with parameter "foo" and second with parameter 0, which is the value of the second bracket. Feels a little bit like this 0 is the wrong thing to assign here, but I'm not sure.

THIS ISSUE HAS BEEN MOVED TO GITHUB https://github.com/dlang/phobos/issues/9980 DO NOT COMMENT HERE ANYMORE, NOBODY WILL SEE IT, THIS ISSUE HAS BEEN MOVED TO GITHUB