DIP29: implicit conversion of pointer expressions

[WalterBright]

This implements implicit conversions of (ptr+offset) expressions so that the ptr part determines if the expression is implicitly convertible.

This is a reboot of #1700

http://wiki.dlang.org/DIP29

[WalterBright]

Just ignore the testCols.d thing, it's just that damned CRLF problem again that git thinks is different but is not.

[9rnsr]

This line will make an invalid pointer. I think we should avoid any undefined behaviors even in trivial test cases, if it's not intended.

[WalterBright]

Done.

[9rnsr]

Note that, CRLF fix for testCols.d is already done #3307 in git head.so just you can remove the commit.

[yebblies]

Should be static?

[yebblies]

Alternatively, change this to return void and use t and result member variables.

[WalterBright]

made it static

[WalterBright]

Note that, CRLF fix for testCols.d is already done #3307 in git head.so just you can remove the commit.

Great, but I have no idea how to do that.

[yebblies]

Great, but I have no idea how to do that.

git checkout master
git pull --ff-only upstream master
git checkout DIP29_1
git rebase -i master
<delete the commit you don't want from the list>
git push -f

[WalterBright]

@yebblies that worked like a champ! You da man!

[yebblies]

Why tail immutable instead of full immutable? The immutableness of the head shouldn't make any difference here...

[WalterBright]

Either way works, but I wished to make it clear that what matters with pointer conversions is what is pointed to, not the pointer itself.

[yebblies]

I don't think this is clearer. Put that note in a comment instead of complicating the code (and making it slower).

[WalterBright]

I should add that this should work like a pure function:

pure T* foo(S* p) { return p + offset; }

[9rnsr]

pure T* foo(S* p) { return p + offset; }

Only when S* is implicitly convertible to T*? Without concrete full code, I cannot see it is correct or not.

[yebblies]

Should have a bugzilla report, not just a DIP.

[WalterBright]

Sorry, should be:

pure T* foo(S* p) { return cast(T*)(p + offset); }

@yebblies , the last time everyone said it should be a dip.

[yebblies]

the last time everyone said it should be a dip.

Having a DIP is great, but you still need a bugzilla entry for the changelog.

[WalterBright]

https://d.puremagic.com/issues/show_bug.cgi?id=12232

[9rnsr]

Looks good.

[9rnsr]

Auto-merge toggled on