On 03.02.2014 19:54, evilrat wrote:
> On Monday, 3 February 2014 at 18:37:20 UTC, Andrew Edwards wrote:
>> Windows users, please give the new installer a try. It has been
>> updated to facilitate proper installation.
>
> vc2013/win 8.1/winsdk 8.1
> detects visual studio correctly, path to win sdk correctly.
>
> has wrong path to mspdb120.dll, current:
> PATH=%PATH%;%VCINSTALLDIR%\bin\x86_amd64;%VCINSTALLDIR%\..\Common7\IDE
> should be:
> PATH=%PATH%;%VCINSTALLDIR%\bin;%VCINSTALLDIR%\..\Common7\IDE

There is an unfortunate collision with the search paths for 64-bit and 32-bit linker:

1. If you let the installer patch sc.ini to use the latest VS installation, it will use the 64-bit linker, and the settings are ok

2. If the installer does patch sc.ini or if you just use the zip-distribution, the 32-bit linker is the default as it has the same installation path for all supported versions of VS.

I guess we could always add the line

PATH=%PATH%;%VCINSTALLDIR%\bin;%VCINSTALLDIR%\..\Common7\IDE

after the VS version specific lines.

>
> has no path to libs dir for sdk 8.1, should be:
> ; Platform libraries (Windows SDK 8.1)
> LIB=%LIB%;"%WindowsSdkDir%\Lib\winv6.3\um\x64"

According to commit log, this has been taking care of.

On Monday, 3 February 2014 at 18:34:15 UTC, Andrew Edwards wrote:
> Following are the changes incorporated since beta 2:

> The list of current regressions may be accessed here:
>
> http://d.puremagic.com/issues/buglist.cgi?query_format=advanced&bug_severity=regression&bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED
>
> Regards,
> Andrew

I just found a disastrous optimizer bug in our production code.
https://d.puremagic.com/issues/show_bug.cgi?id=12095
We shouldn't do a release without fixing that one.

Ouch! Wonder why the auto tester never picked that up.
On 7 Feb 2014 10:40, "Don" <x@nospam.com> wrote:

> On Monday, 3 February 2014 at 18:34:15 UTC, Andrew Edwards wrote:
>
>> Following are the changes incorporated since beta 2:
>>
>
>  The list of current regressions may be accessed here:
>>
>> http://d.puremagic.com/issues/buglist.cgi?query_format= advanced&bug_severity=regression&bug_status=NEW&bug_ status=ASSIGNED&bug_status=REOPENED
>>
>> Regards,
>> Andrew
>>
>
> I just found a disastrous optimizer bug in our production code.
> https://d.puremagic.com/issues/show_bug.cgi?id=12095
> We shouldn't do a release without fixing that one.
>

On Friday, 7 February 2014 at 08:44:34 UTC, Rory McGuire wrote:
> Ouch! Wonder why the auto tester never picked that up.
> On 7 Feb 2014 10:40, "Don" <x@nospam.com> wrote:

Because of no final by default?

On Friday, 7 February 2014 at 10:00:50 UTC, Francesco Cattoglio wrote:
> On Friday, 7 February 2014 at 08:44:34 UTC, Rory McGuire wrote:
>> Ouch! Wonder why the auto tester never picked that up.
>> On 7 Feb 2014 10:40, "Don" <x@nospam.com> wrote:
>
> Because of no final by default?

No. The bug has probably always been present in the 64 bit DMD. Historically, a couple of optimizer bugs like this one have been discovered each year.

On 2/4/14, 2:34, Andrew Edwards wrote:
> Installer
> ...

These are the correct titles for the Installer pull requests:

#52 add alternative download for dmd.${Version2}.zip
#51 disable Audio and USB in VirtualBox
#50 use lib64 for phobo64.lib and gcstub64.obj on Windows
#49 only copy explicitly listed files from phobos
#48 fix deb copyright typo
#47 Update deb/rpm to new dmd versioning scheme.
#46 Vagrant
#45 update symlinks

On 2/7/2014 2:32 AM, Don wrote:
> No. The bug has probably always been present in the 64 bit DMD. Historically, a
> couple of optimizer bugs like this one have been discovered each year.

It's in 32 bit DMD too.

On 02/03/2014 07:34 PM, Andrew Edwards wrote:
> http://ftp.digitalmars.com/libphobos2-65_2.065.0-b3-0_amd64.deb
> http://ftp.digitalmars.com/libphobos2-65_2.065.0-b3-0_i386.deb

Do we need separate libphobos2 debian packages?
Until now I've never seen them on the website (http://dlang.org/download.html) and AFAIK libphobs2.* is already in the dmd package.

El 07/02/14 16:56, Martin Nowak ha escrit:
> On 02/03/2014 07:34 PM, Andrew Edwards wrote:
>> http://ftp.digitalmars.com/libphobos2-65_2.065.0-b3-0_amd64.deb http://ftp.digitalmars.com/libphobos2-65_2.065.0-b3-0_i386.deb
> 
> Do we need separate libphobos2 debian packages?
> Until now I've never seen them on the website (http://dlang.org/download.html) and AFAIK libphobs2.* is already in the dmd package.
> 

That's correct, libphobos2.* is in the all-in-one dmd deb package. If a program compiled against libphobos2.so.*.*.* should be run in a third computer, you need this library on that system.

libphobos2-63, libphobos2-64, etc. only contains the phobos shared library of that version for run-time purposes only.

The version is included on their names allowing to install multiple phobos shared libraries version at same time.

As dmd package contains this library to, the same version of libphobos2-?? and dmd deb packages conflicts, so they cannot be installed together.

Regards,
-- 
Jordi Sayol

On Monday, 3 February 2014 at 19:20:05 UTC, Andrew Edwards wrote:
> On 2/3/14, 2:12 PM, Rory McGuire wrote:
>> http://ftp.digitalmars.com/__dmd_2.065.0-b3-0_amd64.deb is 404 for me.
>
> Linux
>     http://ftp.digitalmars.com/dmd_2.065.0~b3-0_amd64.deb
>     http://ftp.digitalmars.com/dmd_2.065.0~b3-0_i386.deb

First, the work you do is awesome. This release will be much nicer than previous one. Everybody have the option to test easily and report regressions.

As far as I can tell, I have no regression in my code (it compiles and test suite passes).

Good jib everybody. Don's bug is quite scary, so probably want to release a beta 4.