[dub] Release build fails with 2.067.0

John Colvin via Digitalmars-d-learn digitalmars-d-learn at puremagic.com
Wed Apr 15 03:53:35 PDT 2015


On Wednesday, 15 April 2015 at 10:31:01 UTC, Chris wrote:
> On Wednesday, 15 April 2015 at 10:26:55 UTC, John Colvin wrote:
>> On Wednesday, 15 April 2015 at 10:11:17 UTC, Chris wrote:
>>> On Tuesday, 14 April 2015 at 15:58:18 UTC, Chris wrote:
>>>> Sorry, I forgot the title the first time around!
>>>>
>>>> I get the following message from dub when using 
>>>> "--build=release":
>>>>
>>>> Linking...
>>>> .dub/build/myprogram-release-linux.posix-x86_64-dmd_2067-C1A5273464ACE9961E3F3BA6AC04084B/abairtts.o:(.data._D65TypeInfo_xC3std5range10interfaces18__T10InputRangeTiZ10InputRange6__initZ+0x10):
>>>> undefined reference to
>>>> `_D64TypeInfo_C3std5range10interfaces18__T10InputRangeTiZ10InputRange6__initZ'
>>>> collect2: error: ld returned 1 exit status
>>>> --- errorlevel 1
>>>>
>>>> DUB version 0.9.23, built on Apr  3 2015
>>>> dmd 2.067.0
>>>>
>>>> Also, if I set the flags manually (dlfags[1]), they are 
>>>> ignored
>>>> and dub build in debug mode.
>>>>
>>>> [1] now -boundscheck=off, instead of -noboundscheck
>>>
>>> Update:
>>>
>>> I tried to compile the code with a script and I still have 
>>> the same problem:
>>>
>>> dmd -release -boundscheck=off -O -inline
>>> (or any combination of these)
>>>
>>> When I use
>>>
>>> dmd file1.d, file2.d ...
>>>
>>> it also fails. Only when I use
>>>
>>> dmd -debug file1.d file2.d ...
>>>
>>> does it work. Isn't that weird.
>>>
>>> I've made sure that the latest version of dmd is now 
>>> installed system wide (Ubuntu) and not just locally with dvm 
>>> (which is a cool tool). I have no clue what causes this 
>>> problem.
>>
>> How about with -allinst?
>
> This did the trick, thanks a million! But why would I have to 
> do this now (it wasn't necessary with 2.066)?
>
> Also, if you run
>
> dub --build=profile
>
> you'll get
>
> src/phobos/std/getopt.d(438): Warning: statement is not 
> reachable
> src/phobos/std/getopt.d(577): Warning: statement is not 
> reachable
>
> The plot thickens ... :-)

Not a clue, sorry. If you can get minimal test cases for these 
(e.g. with dustmite), bug reports would be good.


More information about the Digitalmars-d-learn mailing list