[phobos] Phobos broken by svn 2233.

[Andrei Alexandrescu]

This looks like a bug in the compiler. This line:

alias typeof(fun(.ElementType!R.init)) ElementType;

causes the error message

std/algorithm.d(118): Error: variable 
std.algorithm.Map!(result,Repeat!(HasPostblit)).Map.__tmp1067 of type 
struct HasPostblit uses this(this), which is not allowed in static 
initialization

but this line:

alias typeof({ return fun(.ElementType!R.init); }()) ElementType;

works. I don't see a reason for which the first should work but the 
second shouldn't. I gave myself five minutes to reproduce the bug in a 
shorter context, without success. (Cue Walter with the line "But it's 
dead easy!") So I'll commit but will leave the non-working code 
commented out along with a @@@BUG@@@ comment in case one of you wants to 
take a look.

Speaking of messages during compilation, what's the deal with this?

Warning: WhiteHole!(I_1) used assert(0) instead of Error for the 
auto-implemented nothrow function I_1.bar
Warning: AutoImplement!(C_6) ignored variadic arguments to the 
constructor C_6(...)


Andrei

On 12/27/10 11:59 PM, Don Clugston wrote:
> If it doesn't happen on OSX, that's very odd, since it shows up in
> Brad's autotester results for both Win32 and Linux32, though in
> different places on each.
> Note that it does NOT stop the build, so it's quite easy to miss, you
> have to scroll through the output. On Linux, it happens after:
> Testing generated/linux/debug/unittest/std/range
>
>
> On 28 December 2010 04:45, Andrei Alexandrescu<andrei at erdani.com>  wrote:
>> Can't reproduce on OSX. I built and unittested both the debug and release
>> versions. Anyone willing to take a look at this issue on Windows?
>>
>> Andrei
>>
>> On 12/27/10 10:14 AM, Don Clugston wrote:
>>>
>>> Compiling with -unittest on all platforms yields:
>>>
>>> std\algorithm.d(118): Error: variable
>>> std.algorithm.Map!(result,Repeat!(HasPostblit)).Map.__tmp5735 of type
>>> struct HasPostblit uses this(this), which is not allowed in static
>>> initialization
>>>
>>> You can see this in the DAutoTester results.
>>>
>>> Introduced in Phobos svn 2233
>>> "Fixed unlisted bug in readf"
>>> in which the line 118 was actually modified. An odd thing about this
>>> error is that the build continues regardless, and the tests still
>>> pass.
>>> _______________________________________________
>>> phobos mailing list
>>> phobos at puremagic.com
>>> http://lists.puremagic.com/mailman/listinfo/phobos
>>
>> _______________________________________________
>> phobos mailing list
>> phobos at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/phobos
>>
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos