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

std.bigint is missing from the sidebar. (I think this happened in the
last release, too).

On 14 September 2010 23:20, Walter Bright <walter at digitalmars.com> wrote:
> http://ftp.digitalmars.com/dmd2beta.zip
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>

  I've put this beta through all my standard tests that I've put the
last few betas through.

What works (i.e. compiles and passes unittests):
Phobos
dstats
gtkD
plot2kill (gtkD port)
std.parallelism  (candidate Phobos lib)
std.rational  (candidate Phobos lib)

Issues:
(Major)  Compiling DFL crashes DMD on dfl's socket.d.  This is related
to changes in Phobos, not DMD.  Compiling DFL with DMD 2.049 + Phobos
2.049 OR DMD 2.048 + SVN Phobos crashes DMD, but compiling w/ DMD 2.048
+ Phobos 2.048 doesn't.  I'm working on figuring out exactly when/how
this broke.  I will keep the list posted.

(Minor)  I found a bug in std.array.array() where it doesn't work with opApply-based iterables with length.  (My bad.)  I've checked in the fix already.  This doesn't break existing code because it's related to new functionality.

On 9/14/2010 5:20 PM, Walter Bright wrote:
> http://ftp.digitalmars.com/dmd2beta.zip
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>

> Issues:
> (Major)  Compiling DFL crashes DMD on dfl's socket.d.  This is related
> to changes in Phobos, not DMD.  Compiling DFL with DMD 2.049 + Phobos
> 2.049 OR DMD 2.048 + SVN Phobos crashes DMD, but compiling w/ DMD
> 2.048 + Phobos 2.048 doesn't.  I'm working on figuring out exactly
> when/how this broke.  I will keep the list posted.
>

Ok, I've done the dirty work of binary searching through revisions to find the one that actually broke this.  It's Phobos revision 1923 (http://dsource.org/projects/phobos/changeset/1923).  If I use DMD 2.049 + druntime 2.049 + Phobos SVN 1922, DFL compiles fine.  If I use DMD 20.49 + druntime 2.049 + Phobos SVN 1923, the compiler just hard crashes without any error message whatsoever on Windows 7.

  I've found the offending lines of code and committed a workaround to
Phobos.  DFL now compiles properly.  Obviously the underlying bug is in
DMD, not Phobos, but I suggest we go ahead w/ this release anyhow because:

1.  I have no freakin' clue how to even reproduce the underlying DMD bug in a reasonably small test case, let alone fix it.  About all I know is that it's somehow related to auto returns (because that's what "solved" the problem when I changed it in Phobos).  My gut feeling is that this is probably one of those bugs that only shows up under a very weird combination of circumstances, such that it's very unlikely anyone will run into it.

2.  It's not a regression (or at least not a recent one) because the bug shows up using 2.048 DMD with recent SVNs of Phobos.

On 9/14/2010 7:55 PM, David Simcha wrote:
>
>> Issues:
>> (Major)  Compiling DFL crashes DMD on dfl's socket.d.  This is
>> related to changes in Phobos, not DMD.  Compiling DFL with DMD 2.049
>> + Phobos 2.049 OR DMD 2.048 + SVN Phobos crashes DMD, but compiling
>> w/ DMD 2.048 + Phobos 2.048 doesn't.  I'm working on figuring out
>> exactly when/how this broke.  I will keep the list posted.
>>
>
> Ok, I've done the dirty work of binary searching through revisions to find the one that actually broke this.  It's Phobos revision 1923 (http://dsource.org/projects/phobos/changeset/1923).  If I use DMD 2.049 + druntime 2.049 + Phobos SVN 1922, DFL compiles fine.  If I use DMD 20.49 + druntime 2.049 + Phobos SVN 1923, the compiler just hard crashes without any error message whatsoever on Windows 7.

New beta up with your corrections.

Hi,

I just noticed that the beta still contains the "old" linker 8.00.2.

(which is kind of good as the mago-debugger has troubles with the new version. It does no longer show watches/locals. The only difference in debug info I could see is that the new version writes udt 'object' twice (not 'object.Object'), but I'm not sure if this triggers the problem. Anything else changed with respect to debug info?)

Rainer

Walter Bright wrote:
> http://ftp.digitalmars.com/dmd2beta.zip
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta
>

On 15 September 2010 06:26, Walter Bright <walter at digitalmars.com> wrote:
> New beta up with your corrections.

This bug was fixed (I think it's important to include in the changelog
-- the submitter had complained about it not being fixed).
3544 optlink termination 0041338f with recursive nested functions

Of course this doesn't need a new beta <g>.

Rainer Schuetze wrote:
>
>
> I just noticed that the beta still contains the "old" linker 8.00.2.

Fixed.

>
> (which is kind of good as the mago-debugger has troubles with the new version. It does no longer show watches/locals. The only difference in debug info I could see is that the new version writes udt 'object' twice (not 'object.Object'), but I'm not sure if this triggers the problem. Anything else changed with respect to debug info?)
>

Hmm, I haven't noticed problems with the debug info with windbg. What's the mago-debugger?

I noticed the difference of isOutputRange behavior.

http://ideone.com/HsmaL

Pointer version of isOutputRange returns false on 2.048 and 2.049beta. I did't find the related changelog. Is this a bug?


Masahiro

On Wed, 15 Sep 2010 06:20:32 +0900, Walter Bright <walter at digitalmars.com> wrote:

> http://ftp.digitalmars.com/dmd2beta.zip
> _______________________________________________
> dmd-beta mailing list
> dmd-beta at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/dmd-beta