> On 5 May 2011 08:25, Jonathan M Davis <jmdavisProg at gmx.com> wrote:
> >> On 5/4/2011 9:17 PM, Jonathan M Davis wrote:
> >> > The one thing that I'm aware of that was a problem thanks to Don's CTFE changes which I don't know whether it was fixed or not was QtD. IIRC, Max was saying that it was seriously broken earlier. There's a good chance that it's fine now, but I think that we should make sure that QtD is no longer failing to build due to the CTFE changes.
> >> 
> >> I folded in patches to dmd requested by Qt. If there are other issues with Qt, are they in bugzilla?
> > 
> > No, I don't believe so. I just recall that when the recent CTFE changes came up in one of the threads in the main newsgroup, and I mentioned that they were still broken, because my recent changes to std.datetime weren't compiling, Max said that they were causing a lot of failures for QtD as well. I doubt that he would have created a bug report though, given that Don was in the middle of his changes. It's quite possible that it compiles fine now though. I just think that it should be verified that the CTFE changes are now stable enough that QtD is no longer failing to build because of them as it was before. Much as Don's changes are definitely something that we want, they risk causing serious regressions if we're not careful.
> 
> I added a lot more very harsh sanity checks to the CTFE engine, so
> there's very little risk of wrong-code regressions.
> If there are any regressions, they will probably be an ICE, on
> something which previously worked by pure luck.
> 
> I think we should roll out a beta, as soon as Phobos unittests pass.

That's fine with me. I don't see a problem with that. I just know that stuff was definitely broken before while you were making your changes, and we need to do our best to make sure that such regressions are fixed or that they shouldn't have been compiling in the first place. Hopefully the beta will catch those.

- Jonathan M Davis

On 5/5/2011 3:49 AM, Don Clugston wrote:
> I think we should roll out a beta, as soon as Phobos unittests pass.

I have one small bug that I want to absolutely insist gets fixed before we even think of releasing.  It's trivial, but I can't fix it now because other things are too broken for me to test my fix:

http://d.puremagic.com/issues/show_bug.cgi?id=5891


Whenever everything else is fixed enough, someone please just change auto length() to typeof(unsigned(pastLast - current)) in the iota implementation, make sure it works and commit it.

I consider this an absolute must fix because having iota not have a length completely cripples a major parallel foreach use case.

On Thu, May 5, 2011 at 9:25 AM, Jonathan M Davis <jmdavisProg at gmx.com>wrote:

> > On 5/4/2011 9:17 PM, Jonathan M Davis wrote:
> > > The one thing that I'm aware of that was a problem thanks to Don's CTFE changes which I don't know whether it was fixed or not was QtD. IIRC,
> Max
> > > was saying that it was seriously broken earlier. There's a good chance that it's fine now, but I think that we should make sure that QtD is no longer failing to build due to the CTFE changes.
> >
> > I folded in patches to dmd requested by Qt. If there are other issues
> with
> > Qt, are they in bugzilla?
>
> No, I don't believe so. I just recall that when the recent CTFE changes
> came
> up in one of the threads in the main newsgroup, and I mentioned that they
> were
> still broken, because my recent changes to std.datetime weren't compiling,
> Max
> said that they were causing a lot of failures for QtD as well.


There is a confusion somewhere. I have been distracted from QtD recently and have not followed the NG either. Unless I sleepwalk, I couldn't try Don's CTFE changes.


> I doubt that he
> would have created a bug report though, given that Don was in the middle of
> his changes. It's quite possible that it compiles fine now though. I just
> think that it should be verified that the CTFE changes are now stable
> enough
> that QtD is no longer failing to build because of them as it was before.
> Much
> as Don's changes are definitely something that we want, they risk causing
> serious regressions if we're not careful.
>

I have just tried to build the public branch of QtD with a recent dmd ( https://github.com/D-Programming-Language/dmd/commit/3fcc2344681ab9d31d1897188f36051c9beb247f) on linux. The build has completed without errors.

http://d.puremagic.com/issues/show_bug.cgi?id=3516 remains critical.


>
> - Jonathan M Davis
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20110505/0dc28a02/attachment.html>

On 5 May 2011 20:14, Max Samukha <maxsamukha at gmail.com> wrote:
>
> On Thu, May 5, 2011 at 9:25 AM, Jonathan M Davis <jmdavisProg at gmx.com> wrote:
>>
>> > On 5/4/2011 9:17 PM, Jonathan M Davis wrote:
>> > > The one thing that I'm aware of that was a problem thanks to Don's
>> > > CTFE
>> > > changes which I don't know whether it was fixed or not was QtD. IIRC,
>> > > Max
>> > > was saying that it was seriously broken earlier. There's a good chance
>> > > that it's fine now, but I think that we should make sure that QtD is
>> > > no
>> > > longer failing to build due to the CTFE changes.
>> >
>> > I folded in patches to dmd requested by Qt. If there are other issues
>> > with
>> > Qt, are they in bugzilla?
>>
>> No, I don't believe so. I just recall that when the recent CTFE changes
>> came
>> up in one of the threads in the main newsgroup, and I mentioned that they
>> were
>> still broken, because my recent changes to std.datetime weren't compiling,
>> Max
>> said that they were causing a lot of failures for QtD as well.
>
>
> There is a confusion somewhere. I have been distracted from QtD recently and have not followed the NG either. Unless I sleepwalk, I couldn't try Don's CTFE changes.
>
>>
>> I doubt that he
>> would have created a bug report though, given that Don was in the middle
>> of
>> his changes. It's quite possible that it compiles fine now though. I just
>> think that it should be verified that the CTFE changes are now stable
>> enough
>> that QtD is no longer failing to build because of them as it was before.
>> Much
>> as Don's changes are definitely something that we want, they risk causing
>> serious regressions if we're not careful.
>
> I have just tried to build the public branch of QtD with a recent dmd (https://github.com/D-Programming-Language/dmd/commit/3fcc2344681ab9d31d1897188f36051c9beb247f) on linux. The build has completed without errors.
>
> http://d.puremagic.com/issues/show_bug.cgi?id=3516 remains critical.

3516 is fixed AFAIK. Do you have a test case that still fails?

> On Thu, May 5, 2011 at 9:25 AM, Jonathan M Davis <jmdavisProg at gmx.com>wrote:
> > > On 5/4/2011 9:17 PM, Jonathan M Davis wrote:
> > > > The one thing that I'm aware of that was a problem thanks to Don's CTFE changes which I don't know whether it was fixed or not was QtD. IIRC,
> > 
> > Max
> > 
> > > > was saying that it was seriously broken earlier. There's a good chance that it's fine now, but I think that we should make sure that QtD is no longer failing to build due to the CTFE changes.
> > > 
> > > I folded in patches to dmd requested by Qt. If there are other issues
> > 
> > with
> > 
> > > Qt, are they in bugzilla?
> > 
> > No, I don't believe so. I just recall that when the recent CTFE changes
> > came
> > up in one of the threads in the main newsgroup, and I mentioned that they
> > were
> > still broken, because my recent changes to std.datetime weren't
> > compiling, Max
> > said that they were causing a lot of failures for QtD as well.
> 
> There is a confusion somewhere. I have been distracted from QtD recently and have not followed the NG either. Unless I sleepwalk, I couldn't try Don's CTFE changes.

Yeah. I mis-remembered. I just checked, and it was David Nadlinger who pointed out that the CTFE changes broke CTFE. So, unless you sleepwalk (sleeptype?) as David Nadlinger, it wasn't you who posted about it. ;)

> > I doubt that he
> > would have created a bug report though, given that Don was in the middle
> > of his changes. It's quite possible that it compiles fine now though. I
> > just think that it should be verified that the CTFE changes are now
> > stable enough
> > that QtD is no longer failing to build because of them as it was before.
> > Much
> > as Don's changes are definitely something that we want, they risk causing
> > serious regressions if we're not careful.
> 
> I have just tried to build the public branch of QtD with a recent dmd ( https://github.com/D-Programming-Language/dmd/commit/3fcc2344681ab9d31d1897 188f36051c9beb247f) on linux. The build has completed without errors.

That's good to hear. Then whatever CTFE changes broke the QtD build were completed/fixed.

- Jonathan M Davis

On Thu, May 5, 2011 at 9:26 PM, Don Clugston <dclugston at googlemail.com>wrote:

> On 5 May 2011 20:14, Max Samukha <maxsamukha at gmail.com> wrote:
> >
> > On Thu, May 5, 2011 at 9:25 AM, Jonathan M Davis <jmdavisProg at gmx.com> wrote:
> >>
> >> > On 5/4/2011 9:17 PM, Jonathan M Davis wrote:
> >> > > The one thing that I'm aware of that was a problem thanks to Don's
> >> > > CTFE
> >> > > changes which I don't know whether it was fixed or not was QtD.
> IIRC,
> >> > > Max
> >> > > was saying that it was seriously broken earlier. There's a good
> chance
> >> > > that it's fine now, but I think that we should make sure that QtD is
> >> > > no
> >> > > longer failing to build due to the CTFE changes.
> >> >
> >> > I folded in patches to dmd requested by Qt. If there are other issues
> >> > with
> >> > Qt, are they in bugzilla?
> >>
> >> No, I don't believe so. I just recall that when the recent CTFE changes
> >> came
> >> up in one of the threads in the main newsgroup, and I mentioned that
> they
> >> were
> >> still broken, because my recent changes to std.datetime weren't
> compiling,
> >> Max
> >> said that they were causing a lot of failures for QtD as well.
> >
> >
> > There is a confusion somewhere. I have been distracted from QtD recently
> and
> > have not followed the NG either. Unless I sleepwalk, I couldn't try Don's CTFE changes.
> >
> >>
> >> I doubt that he
> >> would have created a bug report though, given that Don was in the middle
> >> of
> >> his changes. It's quite possible that it compiles fine now though. I
> just
> >> think that it should be verified that the CTFE changes are now stable
> >> enough
> >> that QtD is no longer failing to build because of them as it was before.
> >> Much
> >> as Don's changes are definitely something that we want, they risk
> causing
> >> serious regressions if we're not careful.
> >
> > I have just tried to build the public branch of QtD with a recent dmd (
> https://github.com/D-Programming-Language/dmd/commit/3fcc2344681ab9d31d1897188f36051c9beb247f )
> > on linux. The build has completed without errors.
> >
> > http://d.puremagic.com/issues/show_bug.cgi?id=3516 remains critical.
>
> 3516 is fixed AFAIK. Do you have a test case that still fails?
>

I don't know. My development branch is in the middle of a big change. It will take some time to complete the change. Then I will remove workarounds for 3516 in the unittests and see whether they all pass.


> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20110505/4aafa1bb/attachment.html>

An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20110505/f439d22a/attachment.html>

On 5/5/2011 6:44 PM, David Simcha wrote:
>
> I've found the extremely subtle race condition that was destroying stack-allocated tasks in amap and parallel foreach while they were still referenced and causing scope closures to be deallocated prematurely in a few ridiculously rare cases.  This is obviously capable of causing stack corruption.  I'm going to try to fix it tonight, but I'm not guaranteeing anything.

Awesome! How'd you find it?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20110505/183429be/attachment.html>

An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20110505/788648e2/attachment.html>

On 5/5/2011 11:26 AM, Don Clugston wrote:
> On 5 May 2011 20:14, Max Samukha <maxsamukha at gmail.com> wrote:
>> http://d.puremagic.com/issues/show_bug.cgi?id=3516 remains critical.
> 
> 3516 is fixed AFAIK. Do you have a test case that still fails?

The specific code in 3516 ought to work now.  However, destructors in the face of exception handling is still not fully solved, and almost certainly won't be for this release.