Files are uploaded to https://dlang.dawg.eu/downloads/dmd.2.067.0-b2/ and will soon be mirrored to http://ftp.digitalmars.com.

We had some troubles with the documentation generation.
In order to not delay the beta even further the zip files don't contain documentation, the dman tool is missing and there are no installers.
We'll try to deliver those subsequently or with the next beta.

Please try out the beta with any code you can compile and report issues at https://issues.dlang.org/.

Once the beta is mirrored to ftp.digitalmars.com, it will also be possible to test on Travis-CI using this line in you .travis.yml.

d: dmd-2.067.0-b2

http://docs.travis-ci.com/user/languages/d/


A yet incomplete changelog can be found here.
http://dlang.org/changelog.html#new2_067

I'd be particularly interested in getting any feedback for the various GC improvements.

Happy Testing,
Martin
_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta

And mirrored to downloads.dlang.org as well.

On 2/17/2015 9:01 PM, Martin Nowak via dmd-beta wrote:
> Files are uploaded to https://dlang.dawg.eu/downloads/dmd.2.067.0-b2/
> and will soon be mirrored to http://ftp.digitalmars.com.
>
> We had some troubles with the documentation generation.
> In order to not delay the beta even further the zip files don't contain
> documentation, the dman tool is missing and there are no installers.
> We'll try to deliver those subsequently or with the next beta.
>
> Please try out the beta with any code you can compile and report issues
> at https://issues.dlang.org/.
>
> Once the beta is mirrored to ftp.digitalmars.com, it will also be
> possible to test on Travis-CI using this line in you .travis.yml.
>
> d: dmd-2.067.0-b2
>
> http://docs.travis-ci.com/user/languages/d/
>
>
> A yet incomplete changelog can be found here.
> http://dlang.org/changelog.html#new2_067
>
> I'd be particularly interested in getting any feedback for the various
> GC improvements.
>
> Happy Testing,
> Martin
> _______________________________________________
> dmd-beta mailing list
> dmd-beta@puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta

Added beta release job to my jenkins box : https://jenkins.dicebot.lv/job/dlang-beta

For now looks like only vibe.d master is not working with it but I will keep adding more projects.

On Wed, Feb 18, 2015 at 8:07 AM, Brad Roberts via dmd-beta < dmd-beta@puremagic.com> wrote:

> And mirrored to downloads.dlang.org as well.
>
>
> On 2/17/2015 9:01 PM, Martin Nowak via dmd-beta wrote:
>
>> Files are uploaded to https://dlang.dawg.eu/downloads/dmd.2.067.0-b2/ and will soon be mirrored to http://ftp.digitalmars.com.
>>
>> We had some troubles with the documentation generation.
>> In order to not delay the beta even further the zip files don't contain
>> documentation, the dman tool is missing and there are no installers.
>> We'll try to deliver those subsequently or with the next beta.
>>
>> Please try out the beta with any code you can compile and report issues at https://issues.dlang.org/.
>>
>> Once the beta is mirrored to ftp.digitalmars.com, it will also be possible to test on Travis-CI using this line in you .travis.yml.
>>
>> d: dmd-2.067.0-b2
>>
>> http://docs.travis-ci.com/user/languages/d/
>>
>>
>> A yet incomplete changelog can be found here. http://dlang.org/changelog.html#new2_067
>>
>> I'd be particularly interested in getting any feedback for the various GC improvements.
>>
>> Happy Testing,
>> Martin
>> _______________________________________________
>> dmd-beta mailing list
>> dmd-beta@puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>>
> _______________________________________________
> dmd-beta mailing list
> dmd-beta@puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>

On 02/18/2015 07:07 AM, Brad Roberts via dmd-beta wrote:
> And mirrored to downloads.dlang.org as well.

Nice, didn't knew about the pre-release page.
Changed the travis-ci script to reflect that.
https://github.com/travis-ci/travis-build/pull/389
_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta

On 02/18/2015 11:15 AM, Михаил Страшун via dmd-beta wrote:
> Added beta release job to my jenkins box :
> https://jenkins.dicebot.lv/job/dlang-beta

Awesome

> For now looks like only vibe.d master is not working with it but I will
> keep adding more projects.

We're working on that.
https://github.com/rejectedsoftware/vibe.d/pull/972

The problems are related to the new std.concurrency.scheduler which
isn't finished both on phobos' amd vibe.d's side.

The other problem is an unfortunately inevitable backwards incompatibility introduced by making synchronized nothrow.
Vibe.d is one of the few programs that inherit from core.sync.*
classes which now have nothrow lock/unlock methods.

This should be understood as a lecture, how bad an idea it is to allow deriving from a class as part of the public API. It creates the strongest possible coupling with little to no chance for ABI compatibility.
_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta

On 02/18/2015 01:35 PM, Martin Nowak via dmd-beta wrote:
>
> The other problem is an unfortunately inevitable backwards
> incompatibility introduced by making synchronized nothrow.

I'll write a changelog entry for that, explaining how to workaround it.
_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta

On 02/18/2015 07:07 AM, Brad Roberts via dmd-beta wrote:
> And mirrored to downloads.dlang.org as well.

http://downloads.dlang.org/pre-releases/2.x/2.067.0/
_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta

On 18 feb 2015, at 13:35, Martin Nowak via dmd-beta <dmd-beta@puremagic.com> wrote:

> The other problem is an unfortunately inevitable backwards incompatibility introduced by making synchronized nothrow.
> Vibe.d is one of the few programs that inherit from core.sync.*
> classes which now have nothrow lock/unlock methods.

Tango has the same problem.

> This should be understood as a lecture, how bad an idea it is to allow deriving from a class as part of the public API. It creates the strongest possible coupling with little to no chance for ABI compatibility.


Yeah. Classes and methods need to be designed to be overridden. There’s a book, Effective Java (similar to Effective C++), which recommends all methods should be final unless explicitly designed to be virtual.

That’s also a reason why, at least in Eclipse code, there are interfaces with a “2” appended because they couldn’t afford to break backwards compatibly by changing an existing interface.

-- 
/Jacob Carlborg


_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta

On 02/18/2015 07:20 PM, Jacob Carlborg wrote:
> On 18 feb 2015, at 13:35, Martin Nowak via dmd-beta <dmd-beta@puremagic.com> wrote:
>
>> The other problem is an unfortunately inevitable backwards incompatibility introduced by making synchronized nothrow.
>> Vibe.d is one of the few programs that inherit from core.sync.*
>> classes which now have nothrow lock/unlock methods.
>
> Tango has the same problem.

:(
How hard is it to make them nothrow?

If they already are logically nothrow, but it's too much effort to mark all dependencies, you can use this simple fix.

static if (__VERSION__ <= 2066)
    scope (failure) assert(0, "unexpected exception");

We can't deprecate throwing IMonitor lock/unlock methods, because you can't overload/inherit a function that only differs by nothrow.
_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta

On 02/18/2015 07:20 PM, Jacob Carlborg wrote:
>
> Yeah. Classes and methods need to be designed to be overridden. There’s a book, Effective Java (similar to Effective C++), which recommends all methods should be final unless explicitly designed to be virtual.

If you care about ABI compatibility on top of API compat, avoid classes and use interfaces instead. Because adding a single field in the base class changes it's size and breaks the ABI.
And obviously classes shouldn't be abused for code reuse.
_______________________________________________
dmd-beta mailing list
dmd-beta@puremagic.com
http://lists.puremagic.com/mailman/listinfo/dmd-beta