At long last! Thanks for all the hard work, everyone!

http://ftp.digitalmars.com/dmd2beta.zip

Test drove this with plot2kill, dstats and a few unreleased libs. Things that break existing code for me:

1.  http://d.puremagic.com/issues/show_bug.cgi?id=5938

2.  http://d.puremagic.com/issues/show_bug.cgi?id=5939

3.  http://d.puremagic.com/issues/show_bug.cgi?id=5940

The last two have to do with std.algorithm.map being made a nested struct and nested structs having annoying and surprising limitations that the old implementation didn't have.  IMHO this change should be reverted if the compiler isn't going to get fixed in short order.  I guess I'd say the same about Filter and anything else that was recently nested.

Lastly, std.parallelism and std.net.isemail should be listed in the side bar where all the Phobos modules are listed.

On 5/7/2011 12:47 AM, Walter Bright wrote:
> At long last! Thanks for all the hard work, everyone!
>
> http://ftp.digitalmars.com/dmd2beta.zip
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>

On 5/6/2011 10:47 PM, David Simcha wrote:
> Test drove this with plot2kill, dstats and a few unreleased libs.  Things that break existing code for me:
>
> 1.  http://d.puremagic.com/issues/show_bug.cgi?id=5938
>
> 2.  http://d.puremagic.com/issues/show_bug.cgi?id=5939
>
> 3.  http://d.puremagic.com/issues/show_bug.cgi?id=5940
>
> The last two have to do with std.algorithm.map being made a nested struct and nested structs having annoying and surprising limitations that the old implementation didn't have.

This can be fixed by making those nested struct static.

>   IMHO this change should be reverted if the compiler isn't going to get fixed
> in short order.  I guess I'd say the same about Filter and anything else that
> was recently nested.
>
> Lastly, std.parallelism and std.net.isemail should be listed in the side bar where all the Phobos modules are listed.
>
> On 5/7/2011 12:47 AM, Walter Bright wrote:
>> At long last! Thanks for all the hard work, everyone!
>>
>> http://ftp.digitalmars.com/dmd2beta.zip
>> _______________________________________________
>> dmd-beta mailing list
>> dmd-beta at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>>
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>
>

I get this now, although I'm not using to/parse to convert to a floating point type anywhere:

C:\D\dmd2\windows\bin\..\..\src\phobos\std\conv.d(1301): Error: function
std.conv.parse!(real,string).parse compiler error, parameter 'p',
bugzilla 2962?
Assertion failure: '0' on line 744 in file 'glue.c'

With 2.052 I just got it when using parse!double under some obscure circumstances (was not really possible to strip down to a test case).

Also it seems like the linker now strips out less unused functions (if at all). I have to link against external libraries that are used in some modules which are in turn not used anywhere. In all previous versions this code was stripped out and the external depedency was removed.

If compile one of my LIBs using the new dmd i get this:

Error: multiple definition of datetime_364_bdb: _D70C:\_APPS_\_D_\dmd\dmd2\windows\bin\..\..\src\phobos\std\datetime.d.86812__ModuleInfoZ and datetime: _D70C:\_APPS_\_D_\dmd\dmd2\windows\bin\..\..\src\phobos\std\datetime.d.86812__ModuleInfoZ

i looked for it, but i don't even import the phobos datetime module anywhere in this lib.

It just happens with "-release -noboundscheck -inline".

Any ideas ?


On 07.05.2011 06:47, Walter Bright wrote:
> At long last! Thanks for all the hard work, everyone!
>
> http://ftp.digitalmars.com/dmd2beta.zip
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>

On 5/7/2011 2:27 AM, Walter Bright wrote:
>
>
> On 5/6/2011 10:47 PM, David Simcha wrote:
>> Test drove this with plot2kill, dstats and a few unreleased libs. Things that break existing code for me:
>>
>> 1.  http://d.puremagic.com/issues/show_bug.cgi?id=5938
>>
>> 2.  http://d.puremagic.com/issues/show_bug.cgi?id=5939
>>
>> 3.  http://d.puremagic.com/issues/show_bug.cgi?id=5940
>>
>> The last two have to do with std.algorithm.map being made a nested struct and nested structs having annoying and surprising limitations that the old implementation didn't have.
>
> This can be fixed by making those nested struct static.

Indeed this does fix all the problems, but I can't run the Phobos unit tests and commit this fix because I'm hung up on the following druntime build error on Windows:

dmc -c  src\rt\monitor.c
Fatal error: unable to open input file 'mars.h'

On 5/7/2011 10:38 AM, David Simcha wrote:
> On 5/7/2011 2:27 AM, Walter Bright wrote:
>>
>>
>> On 5/6/2011 10:47 PM, David Simcha wrote:
>>> Test drove this with plot2kill, dstats and a few unreleased libs. Things that break existing code for me:
>>>
>>> 1.  http://d.puremagic.com/issues/show_bug.cgi?id=5938
>>>
>>> 2.  http://d.puremagic.com/issues/show_bug.cgi?id=5939
>>>
>>> 3.  http://d.puremagic.com/issues/show_bug.cgi?id=5940
>>>
>>> The last two have to do with std.algorithm.map being made a nested struct and nested structs having annoying and surprising limitations that the old implementation didn't have.
>>
>> This can be fixed by making those nested struct static.
>
> Indeed this does fix all the problems, but I can't run the Phobos unit tests and commit this fix because I'm hung up on the following druntime build error on Windows:
>
> dmc -c  src\rt\monitor.c
> Fatal error: unable to open input file 'mars.h'
>
...and when I try to run the tests on Linux instead I get Error: cannot read file std/net/isemail.d.

It appears that, in your zip package, net ended up as a file instead of a directory.  Its contents are the source code to std.net.isemail.

On 5/7/2011 2:27 AM, Walter Bright wrote:
>
>
> On 5/6/2011 10:47 PM, David Simcha wrote:
>> Test drove this with plot2kill, dstats and a few unreleased libs. Things that break existing code for me:
>>
>> 1.  http://d.puremagic.com/issues/show_bug.cgi?id=5938
>>
>> 2.  http://d.puremagic.com/issues/show_bug.cgi?id=5939
>>
>> 3.  http://d.puremagic.com/issues/show_bug.cgi?id=5940
>>
>> The last two have to do with std.algorithm.map being made a nested struct and nested structs having annoying and surprising limitations that the old implementation didn't have.
>
> This can be fixed by making those nested struct static.

Worked around all the errors I was getting and unfortunately just making them static leads to other errors.  Running the std.algorithm unit tests with the Result struct of Map made static:

std/algorithm.d(388): Error: function
std.algorithm.__unittest11.map!(__dgliteral1).map!(int[]).map.Result.back cannot
get frame pointer to __dgliteral1
std/algorithm.d(422): Error: function
std.algorithm.__unittest11.map!(__dgliteral1).map!(int[]).map.Result.front
cannot get frame pointer to __dgliteral1
std/algorithm.d(429): Error: function
std.algorithm.__unittest11.map!(__dgliteral1).map!(int[]).map.Result.opIndex
cannot get frame pointer to __dgliteral1

> Fatal error: unable to open input file 'mars.h'
>

I'll fix. In the meantime, you can use mars.h from an older release.

> ...and when I try to run the tests on Linux instead I get Error: cannot read file std/net/isemail.d.
>
> It appears that, in your zip package, net ended up as a file instead of a directory.  Its contents are the source code to std.net.isemail.
>

Fixed.