Supplemental changes of dmd/pull/41

[9rnsr]

dlang/dmd#41 makes struct literal and struct constructor call as rvalue, and passing them to ref parameter make invalid.
This pull request adds function overloads for ref and non-ref.

CAUTION: It is necessary to apply changes to phobos and dmd at the same time.

[jmdavis]

As I understand it, you don't actually need the Tdummy=void. Supposedly, () is enough to templatize the function (I haven't tried it though; I just know that David was doing it in some code). Assuming that that works, it would likely be preferable to having Tdummy.

[9rnsr]

Sorry for the late reply. I had missed your comment.

I also think that Tdummy=void is not needed. It might be the garbage of past template limitation.
I couldn't judge it, so keep codes as far as possible.

[andralex]

Ping? Please rebase so I can pull safely. Thanks!

[9rnsr]

Unfortunately, we can't merge this until merging dmd/pull/41.
Because current dmd has a bug that the function overloading ref and non-ref does not work.

  bool opEquals(ref const SList rhs) const { ... }
+ bool opEquals(const SList rhs) const { ... }  // this addition reveals the dmd bug

Then, it is necessary to apply changes to phobos and dmd at the same time.

[andralex]

It would be great to fix this issue in the coming release. Kenji, could you please submit the bug? Walter, any chance we get a fix in the current beta? Thanks!

[9rnsr]

I have already filed the bug as bug 5889, but Walter has a negative opinion against fixing it...

[andralex]

@k-hara: This pull as is right now can be merged independently of dmd changes?

[andralex]

I mean @9rnsr :o)

[9rnsr]

@andralex: Still can't, unfortunately. We MUST merge dmd/pull/41 before this, to avoid conflict of ref/non-ref overload.

[andralex]

@WalterBright: ping

[WalterBright]

Please also fix opEquals() to be @safe or @trusted.

[WalterBright]

Fails miserably at compiling phobos.

[WalterBright]

std\file.d(2103): Error: overloads const pure nothrow bool(const(SysTime) rhs) and const pure nothrow bool(ref const(SysTime) rhs) both match argument list for opEquals
std\file.d(2103): Error: function std.datetime.SysTime.opEquals called with argument types:
((const(SysTime)) const)
matches both:
std.datetime.SysTime.opEquals(const(SysTime) rhs)
and:
std.datetime.SysTime.opEquals(ref const(SysTime) rhs)
std\file.d(2103): Error: overloads const pure nothrow bool(const(SysTime) rhs) and const pure nothrow bool(ref const(SysTime) rhs) both match argument list for opEquals
std\file.d(2103): Error: function std.datetime.SysTime.opEquals called with argument types:
((const(SysTime)) const)
matches both:
std.datetime.SysTime.opEquals(const(SysTime) rhs)
and:
std.datetime.SysTime.opEquals(ref const(SysTime) rhs)
std\file.d(2103): Error: overloads const pure nothrow bool(const(SysTime) rhs) and const pure nothrow bool(ref const(SysTime) rhs) both match argument list for opEquals
std\file.d(2103): Error: function std.datetime.SysTime.opEquals called with argument types:
((const(SysTime)) const)
matches both:
std.datetime.SysTime.opEquals(const(SysTime) rhs)
and:
std.datetime.SysTime.opEquals(ref const(SysTime) rhs)

[9rnsr]

@WalterBright, please read the summary. This pull is an additional fix of dmd/pull/41.
Before ref/non-ref overload issue (it will be fixed by dmd/pull/41), we MUST not merge only this pull.
@Andarex and I argue that you should merge dmd/pull/41 first to merge this pull.

[9rnsr]

Please also fix opEquals() to be @safe or @trusted.

BigInt, SList, and Array are struct, not class. Adding such attribute is better, but it's not related to the purpose of this pull.

[9rnsr]

@WalterBright, Can you reopen this pull? If it isn't possible, I'll post a new pull exactly the same as this.

[WalterBright]

I did pull 41 first. It does not compile phobos.

[9rnsr]

Ouch! Sorry, dmd/pull/41 was silently broken by others pull. And sorry my irrelevant comment (I didn't noticed you already merged the two pulls).
The problem is here. But I cant post a fixup pull while a few hours.

[WalterBright]

reverted for the moment, as well as pull 41.

[yebblies]

@WalterBright Could you please push the reverts then?

[9rnsr]

I've posted a revert pull that contains a small fixup. Please merge #753.