[Issue 16521] Wrong code generation with switch + static foreach

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

--- Comment #13 from Steven Schveighoffer  ---
(In reply to Imperatorn from comment #12)
> Any update on this?

What do you mean? It's resolved as invalid. The code no longer compiles (for a
while now).

--

[Issue 16521] Wrong code generation with switch + static foreach

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

Imperatorn  changed:

   What|Removed |Added

 CC||johan_forsberg_86@hotmail.c
   ||om

--- Comment #12 from Imperatorn  ---
Any update on this?

--

[Issue 16521] Wrong code generation with switch + static foreach

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

LucienPe  changed:

   What|Removed |Added

 Status|NEW |RESOLVED
 CC||lucien.perreg...@gmail.com
 Resolution|--- |INVALID

--

[Issue 16521] Wrong code generation with switch + static foreach

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

Seb  changed:

   What|Removed |Added

 CC||greensunn...@gmail.com

--- Comment #11 from Seb  ---
FWIW this works with static foreach


---
void main ()
{
sformat(0, uint.max);
}

void sformat (Args...) (int index, Args args)
{
JT: switch (index)
{
static foreach (idx, unused; args)
{
case idx:
assert(unused == args[idx], "Borken compiler");
break JT;
}

default:
assert(0);
}
}
---

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

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

--- Comment #10 from Steven Schveighoffer  ---
*** Issue 17218 has been marked as a duplicate of this issue. ***

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

Alex Goltman  changed:

   What|Removed |Added

 CC||alex.golt...@gmail.com

--- Comment #9 from Alex Goltman  ---
looks like a duplicate of my issue
https://issues.dlang.org/show_bug.cgi?id=17218

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

--- Comment #8 from Steven Schveighoffer  ---
guys, the static foreach comment was a *wish* for this, not a recommendation to
use some already existing thing.

It would be nice to do:

static foreach(idx; 0 .. args.length)

Instead of the goofy "unused" symbol thing. In my code, I'm using Args, not
args, so I have no idea why a variable for each tuple item should even need to
be created. It might even be a type!

Note that the "deprecation" is what I'm concerned about, as I have only this
one way to make my switch statement out of a tuple (sorry Stefan, but string
mixin is a *vastly inferior* option compared to this mechanism), and some
future version of dmd I worry is going to flag this as an error, even though I
never use the variable I didn't want to declare but was forced to.

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

--- Comment #7 from Nemanja Boric <4bur...@gmail.com> ---
I'm not sure who ever got convinced by this code that `static foreach` works.
This is just very helpful and clear approach, and it should grant `static
foreach` becoming a real thing, because it is useful.

While that is still not implemented, this code should work, as I see this as a
bug/regression
(https://github.com/rejectedsoftware/vibe.d/blob/6f37e694cc77063769bc4c9a42160627103e8354/web/vibe/web/rest.d#L1367
- yet another example, plus dozens on forums, etc).

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

--- Comment #6 from Stefan Koch  ---
Just because something is used in phobos does not it is correct.
this code is highly dubious because it goats people into believing that tuple
foreach actually works.

The right way to do this is to build a string for the switch and mix it in.

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

Nemanja Boric <4bur...@gmail.com> changed:

   What|Removed |Added

   Keywords|accepts-invalid |

--- Comment #5 from Nemanja Boric <4bur...@gmail.com> ---
I don't agree that this code should error. This is a well known and common D
idiom (simple GH search shows examples in Phobos, such:
https://github.com/dlang/phobos/blob/master/std/algorithm/sorting.d#L1101), up
to the point that there are also merged switch/foreach loops constructs where
the label is applied to switch, so the `break` inside foreach would be
considered switch break.

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

Stefan Koch  changed:

   What|Removed |Added

   Keywords||accepts-invalid
 CC||uplink.co...@gmail.com

--- Comment #4 from Stefan Koch  ---
There is no such thing as static foreach.
You are using tuple foreach which will force an unrolled loop.

The provided code should error!

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

Steven Schveighoffer  changed:

   What|Removed |Added

 CC||schvei...@yahoo.com

--- Comment #3 from Steven Schveighoffer  ---
I see the same deprecation warning in my code, but I'm not using the actual
tuple value, just the index.

Static foreach would be handy here...

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

Nemanja Boric <4bur...@gmail.com> changed:

   What|Removed |Added

 CC||4bur...@gmail.com

--- Comment #2 from Nemanja Boric <4bur...@gmail.com> ---
I hit the same bug today:

```
import std.traits;

struct S
{
int a, b, c, d, e, f;
}

void main()
{
S s;

foreach (name; ["a", "b", "c"])
{
switch (name)
{
foreach (i, ref field; s.tupleof)
{
case __traits(identifier, S.tupleof[i]):
field = 0; // s.tupleof[i] = 0; works
break;
}
default:
break;
}
}
}
```

--

[Issue 16521] Wrong code generation with switch + static foreach

[via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=16521

--- Comment #1 from Mathias Lang  ---
A bit of debugging proved that the following code is generated:

```
[{
enum ulong idx = 0;
uint unused = _param_1;
case idx:
{
assert(unused == args[idx], "Borken compiler");
break JT;
}
}
]
```

Note: Using `ref` results in the program segfaulting because it will then
dereference an uninitialized `ref`.

--