Next release: 2.7.0

[Daniel Lee]

Hi, I don't think there's anything critical that we're waiting on now.

Ben, can you see if the current dev branch is now ok with RStan and all items dependencies?

Daniel

[Daniel Lee]

Oh yeah -- everyone ok with bumping to 2.7.0?

[Ben Goodrich]

Yes, it has been working since the Boost upgrade.

[Daniel Lee]

Great. We'll take inventory today and have an idea of when we can release.

Daniel

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Daniel Lee]

Ben, could you confirm that this commit hash will work for RStan? 822ab42d07db58e617e5682a0ec4013626ba9c87

We should try to install the versions on all 3 operating systems. If you can distribute the package, I'll try on Windows and Mac.

If it works, I will tag that commit hash as 2.7.0. If not, please let me know what needs to be fixed in order for us to tag it.

Daniel

[Michael Betancourt]

Getting the new ibeta derivatives into RStan 2.7.0 would

be immensely helpful to some research efforts of mine...

[Ben Goodrich]

That's fine. StanHeaders is currently pointed to develop and that is the latest commit.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Daniel Lee]

Ben, I don't understand. Have you successfully installed RStan on Linux, Mac, and Windows using that exact version?

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Ben Goodrich]

Not Mac. You can generate Mac binaries from the attachments.

Ben

On Monday, June 1, 2015 at 5:23:59 PM UTC-4, Daniel Lee wrote:

Ben, I don't understand. Have you successfully installed RStan on Linux, Mac, and Windows using that exact version?

On Mon, Jun 1, 2015 at 5:22 PM, Ben Goodrich <goodri...@gmail.com> wrote:

That's fine. StanHeaders is currently pointed to develop and that is the latest commit.

On Monday, June 1, 2015 at 5:13:28 PM UTC-4, Daniel Lee wrote:

Ben, could you confirm that this commit hash will work for RStan? 822ab42d07db58e617e5682a0ec4013626ba9c87

We should try to install the versions on all 3 operating systems. If you can distribute the package, I'll try on Windows and Mac.

If it works, I will tag that commit hash as 2.7.0. If not, please let me know what needs to be fixed in order for us to tag it.

Daniel

On Mon, Jun 1, 2015 at 11:33 AM, Daniel Lee <bea...@alum.mit.edu> wrote:

Great. We'll take inventory today and have an idea of when we can release.

Daniel

On Mon, Jun 1, 2015 at 11:30 AM, Ben Goodrich <goodri...@gmail.com> wrote:

Yes, it has been working since the Boost upgrade.

On Monday, June 1, 2015 at 10:59:59 AM UTC-4, Daniel Lee wrote:

Hi, I don't think there's anything critical that we're waiting on now.

Ben, can you see if the current dev branch is now ok with RStan and all items dependencies?

Daniel

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Jiqiang Guo]

It’s fine on recent Mac, but not using gcc 4.1.2. 

                from file4ea8ab546e.cpp:7:

/Library/Frameworks/R.framework/Versions/3.1/Resources/library/StanHeaders/include/stan/math/prim/scal/prob/poisson_rng.hpp:23: error: call of overloaded 'pow(int, int)' is ambiguous

/usr/include/architecture/i386/math.h:343: note: candidates are: double pow(double, double)

/usr/include/c++/4.2.1/cmath:357: note:                 float std::pow(float, float)

/usr/include/c++/4.2.1/cmath:361: note:                 long double std::pow(long double, long double)

/usr/include/c++/4.2.1/cmath:365: note:                 double std::pow(double, int)

/usr/include/c++/4.2.1/cmath:369: note:                 float std::pow(float, int)

/usr/include/c++/4.2.1/cmath:373: note:                 long double std::pow(long double, int)

make[1]: *** [file4ea8ab546e.o] Error 1

Error: processing vignette 'rstan_vignette.Rnw' failed with diagnostics:

 chunk 3 (label = testify)

Error in Rcpp::sourceCpp(code = paste(lines, collapse = "\n")) :

  Error 1 occurred building shared library.

Execution halted

make: *** [rstan_2.6.2.tar.gz] Error 1

Jiqiang 

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

<StanHeaders_2.7.0.tar.gz><rstan_2.7.0.tar.gz>

[Jiqiang Guo]

Meant to type it was not working using gcc 4.2.1 on old Mac, say, Snow Leopard. 

One way, changing the line has std::pow(2, 32) to std::pow(2.0, 32), solves the problem.  So I think we should fix it since it’s easy.

Jiqiang 

[Bob Carpenter]

I created an issue and assigned it to myself to
carry out Jiqiang's proposed fix for 2.7.0:

https://github.com/stan-dev/stan/issues/1472

Jiqiang --- was everything OK after this change? If there
are other issues like this, please dump into the above issue.

- Bob

[Joshua N Pritikin]

On Mon, Jun 01, 2015 at 02:22:05PM -0700, Ben Goodrich wrote:
> That's fine. StanHeaders is currently pointed to develop and that is
> the latest commit.

When will the StanHeaders on CRAN get updated? Some vigilance is needed
because there are API changes that will require updates to OpenMx.

--
Joshua N. Pritikin
Department of Psychology
University of Virginia
485 McCormick Rd, Gilmer Hall Room 102
Charlottesville, VA 22904
http://people.virginia.edu/~jnp3bc

[Jiqiang Guo]

Thanks.  I think that's it for rstan.

Jiqiang 

[Ben Goodrich]

On Monday, June 1, 2015 at 10:35:40 PM UTC-4, Joshua Pritikin wrote:

On Mon, Jun 01, 2015 at 02:22:05PM -0700, Ben Goodrich wrote:
>    That's fine. StanHeaders is currently pointed to develop and that is
>    the latest commit.

When will the StanHeaders on CRAN get updated? Some vigilance is needed
because there are API changes that will require updates to OpenMx.

Looking like Wednesday.

Ben

[Daniel Lee]

We have our Stan release candidate: 1e38500c045f186a95fa55fdfb873df483305c22

Ben and Jiqiang, can you make sure RStan installs on Linux, Windows and old Mac using that.

Ben, can you send me a version that's packaged to try to install on current Mac.

Daniel, make sure CmdStan installs on Mac and Windows.

If all three of those things are ok, then I'll tag and release.

[Daniel Lee]

I've just confirmed CmdStan installs and runs on Windows and Mac using that commit hash.

[Ben Goodrich]

I think we're good for StanHeaders but let's put in the fix for dirichlet_rng .

[Daniel Lee]

Ben, can you confirm and send me the package to test on Mac? There is no current fix for direchlet_rng, so we won't wait for it. I think it's much higher priority to fix the install issue. 

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Bob Carpenter]

I don't think we should wait for Dirichlet RNG.

It's just a numerical issue with very extreme values
where the draw is on the boundary with a single 1 and
the rest of the simplex 0.

I do have the fix ready to go, I think.

- Bob

> On Jun 2, 2015, at 6:13 PM, Ben Goodrich <goodri...@gmail.com> wrote:
>
> I think we're good for StanHeaders but let's put in the fix for dirichlet_rng .
>
> On Tuesday, June 2, 2015 at 4:53:43 PM UTC-4, Daniel Lee wrote:
> We have our Stan release candidate: 1e38500c045f186a95fa55fdfb873df483305c22
>
> Ben and Jiqiang, can you make sure RStan installs on Linux, Windows and old Mac using that.
> Ben, can you send me a version that's packaged to try to install on current Mac.
> Daniel, make sure CmdStan installs on Mac and Windows.
>
> If all three of those things are ok, then I'll tag and release.
>
>

> On Mon, Jun 1, 2015 at 11:41 PM, Ben Goodrich <goodri...@gmail.com> wrote:
> On Monday, June 1, 2015 at 10:35:40 PM UTC-4, Joshua Pritikin wrote:
> On Mon, Jun 01, 2015 at 02:22:05PM -0700, Ben Goodrich wrote:
> > That's fine. StanHeaders is currently pointed to develop and that is
> > the latest commit.
>
> When will the StanHeaders on CRAN get updated? Some vigilance is needed
> because there are API changes that will require updates to OpenMx.
>
> Looking like Wednesday.
>
> Ben
>
>
>

> --
> You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Ben Goodrich]

Here you go. The install issue would have been fixed days ago when I pushed a fixed install.R if the website would self-update.

Anyway, we have another problem in that the dogs example from BUGS does not sample when compiled with g++-4.9 although it does with clang++-3.6. We're going to need 2.7 to last a while, so we might as well fix known severe issues.

Ben

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Daniel Lee]

Ben, there's a Jenkins job that's manual. You just need to kick it off once logged in. Or you could have let me know. I didn't know that was blocking anything. 

Let me push the website now. Is there an open issue for that problem?

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Bob Carpenter]

Is the reason it needs to last related to CRAN? There's no
reason this is the one that needs to go to CRAN.

You mean mc-stan.org? You need to login and update Git.
We should probably be hosting this through git pages to make
it more integrated. I'm happy to give whoever wants it logins
if they're interested in doing this.

I don't think the Dirichlet issue is severe.

What's the issue with the dogs example? I don't mind fixing
bugs, but at some point we just have to call it. Is this something
major?

- Bob

> On Jun 2, 2015, at 8:27 PM, Ben Goodrich <goodri...@gmail.com> wrote:
>
> Here you go. The install issue would have been fixed days ago when I pushed a fixed install.R if the website would self-update.
>
> Anyway, we have another problem in that the dogs example from BUGS does not sample when compiled with g++-4.9 although it does with clang++-3.6. We're going to need 2.7 to last a while, so we might as well fix known severe issues.
>
> Ben
>
> On Tuesday, June 2, 2015 at 8:20:09 PM UTC-4, Daniel Lee wrote:
> Ben, can you confirm and send me the package to test on Mac? There is no current fix for direchlet_rng, so we won't wait for it. I think it's much higher priority to fix the install issue.
>
> On Tuesday, June 2, 2015, Ben Goodrich <goodri...@gmail.com> wrote:
> I think we're good for StanHeaders but let's put in the fix for dirichlet_rng .
>
> On Tuesday, June 2, 2015 at 4:53:43 PM UTC-4, Daniel Lee wrote:
> We have our Stan release candidate: 1e38500c045f186a95fa55fdfb873df483305c22
>
> Ben and Jiqiang, can you make sure RStan installs on Linux, Windows and old Mac using that.
> Ben, can you send me a version that's packaged to try to install on current Mac.
> Daniel, make sure CmdStan installs on Mac and Windows.
>
> If all three of those things are ok, then I'll tag and release.
>
>

> On Mon, Jun 1, 2015 at 11:41 PM, Ben Goodrich <goodri...@gmail.com> wrote:
> On Monday, June 1, 2015 at 10:35:40 PM UTC-4, Joshua Pritikin wrote:
> On Mon, Jun 01, 2015 at 02:22:05PM -0700, Ben Goodrich wrote:
> > That's fine. StanHeaders is currently pointed to develop and that is
> > the latest commit.
>
> When will the StanHeaders on CRAN get updated? Some vigilance is needed
> because there are API changes that will require updates to OpenMx.
>
> Looking like Wednesday.
>
> Ben
>
>
>
> --
> You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

> For more options, visit https://groups.google.com/d/optout.
>

> --
> You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

> For more options, visit https://groups.google.com/d/optout.

> <StanHeaders_2.7.0.tar.gz><rstan_2.7.0.tar.gz>

[Daniel Lee]

After thinking a bit more, I don't think g++ 4.9 should stop us unless it's distributed as the default compiler for a common Linux distribution. Is it? Mac is on 4.2 and Windows is 4.6.

Jiqiang, can you install that version on your older Mac?

[Alp Kucukelbir]

sorry to jump in late into the convo here. (nips...)

any chance we can squeeze advi into 2.7.0? i think all we're waiting on is documentation (cmdstan and stan manual), which i can provide by wednesday june 10. i also want to adjust some of the default parameters in advi and change some function names. (minor changes.)

what do you think?

cheers
alp

[Ben Goodrich]

There's no GitHub issue. I just changed install.R in the webpages repo to install the previous version of BH.

Ben

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Daniel Lee]

Alp, no. 2.7.0 is going out sooner than that. In addition to what you listed, it still seg faults when using an empty stream. It's definitely not ready right now, so it's not going in.

Daniel

[Ben Goodrich]

On Tuesday, June 2, 2015 at 8:45:31 PM UTC-4, Daniel Lee wrote:

After thinking a bit more, I don't think g++ 4.9 should stop us unless it's distributed as the default compiler for a common Linux distribution. Is it? Mac is on 4.2 and Windows is 4.6.

It's the default CRAN compiler for Linux. The concern here is that I don't yet know what is causing the initialization problem so I don't know how extensive a problem it is.

Ben

[Alp Kucukelbir]

> Alp, no. 2.7.0 is going out sooner than that. In addition to what you listed, it still seg faults when using an empty stream. It's definitely not ready right now, so it's not going in.

roger that. i'm not sure what the empty streams issue is, but let's touch base post 2.7.0 and post nips.

cheers
alp

[Daniel Lee]

Ok. I'll push that now.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Daniel Lee]

Regarding issue -- I meant issue for the dogs example.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Jiqiang Guo]

I tried on Snow Leopard; it’s fine. 

Jiqiang 

[Ben Goodrich]

It affects g++-4.8 too although g++-4.7 is fine. I'll look into it more after I get back from the gym.

Ben

[Daniel Lee]

Ben, the package you sent installs, but with a warning:

In file included from /Library/Frameworks/R.framework/Versions/3.2/Resources/library/BH/include/boost/accumulators/statistics_fwd.hpp:12:
/Library/Frameworks/R.framework/Versions/3.2/Resources/library/BH/include/boost/mpl/print.hpp:50:19: warning:
      in-class initialization of non-static data member is a C++11 extension [-Wc++11-extensions]
    const int m_x = 1 / (sizeof(T) - sizeof(T));
                  ^
1 warning generated.

Do we have controls for includes? If so, can we change:
   -I"/Library/Frameworks/R.framework/Versions/3.2/Resources/library/BH/include"

to

   -isystem"/Library/Frameworks/R.framework/Versions/3.2/Resources/library/BH/include"
?

And maybe apply that to the other libraries too?

[Ben Goodrich]

That warning appears on Linux too but only with clang++, so it is not a problem from CRAN's perspective. Also, it is a BH warning rather than one of ours. I'm surprised Boost released 1.58 with that, but that line of code

#if defined(__clang__)

    const int m_x = 1 / (sizeof(T) - sizeof(T));

is presumably known to be fine with clang even without -std=c++11.

Ben

[Daniel Lee]

I understand that the warning is only a warning. I was asking if we had enough control through RStan to suppress the warning. If so, I think we should suppress warnings from the libraries we use.

[Ben Goodrich]

I'll look into it more. I would have thought plugin.R would cause it to use -isystem but apparently not.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Daniel Lee]

I don't know where in the chain those options get in, or else I would have changed it.

I've got gcc 4.9 on my Mac and trying to install the package you sent earlier. I'll let you know if I have trouble with the dogs example.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Ben Goodrich]

g++-4.9 is putting garbage into xa and xs for some reason

xa =
[0,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630,-2147483629,-2147483628]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483647,-2147483647,-2147483647,-2147483647,-2147483647,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483646,-2147483645,-2147483644,-2147483644,-2147483644,-2147483643,-2147483642,-2147483642,-2147483641,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634]
[0,-2147483648,-2147483647,-2147483646,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483642,-2147483641,-2147483641,-2147483640,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483644,-2147483643,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483647,-2147483647,-2147483647,-2147483647,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483646,-2147483645,-2147483644,-2147483644,-2147483643,-2147483643,-2147483643,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483638,-2147483637,-2147483636]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483647,-2147483646,-2147483645,-2147483645,-2147483644,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631]
[0,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483646,-2147483645,-2147483645,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483645,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630]
[0,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483645,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630,-2147483629]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631]
[0,-2147483648,-2147483647,-2147483647,-2147483646,-2147483646,-2147483646,-2147483646,-2147483645,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631]
[0,-2147483648,-2147483647,-2147483647,-2147483647,-2147483647,-2147483647,-2147483646,-2147483646,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483646,-2147483645,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632]
[0,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630,-2147483629,-2147483628]
[0,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483636,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483645,-2147483644,-2147483644,-2147483644,-2147483644,-2147483643,-2147483642,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633]
[0,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483643,-2147483642,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630,-2147483629]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630,-2147483629]
[0,-2147483648,-2147483648,-2147483648,-2147483647,-2147483647,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483641,-2147483641,-2147483640,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634]
[0,-2147483648,-2147483648,-2147483648,-2147483648,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483642,-2147483641,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483633,-2147483632,-2147483631,-2147483630]
xs =
[0,-2147483647,-2147483646,-2147483646,-2147483645,-2147483645,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483636,-2147483635,-2147483634,-2147483634,-2147483634,-2147483634,-2147483634,-2147483634,-2147483634,-2147483634,-2147483634,-2147483634]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483643,-2147483642,-2147483642,-2147483642,-2147483641,-2147483640,-2147483640,-2147483640,-2147483639,-2147483639,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638]
[0,-2147483647,-2147483647,-2147483647,-2147483646,-2147483645,-2147483645,-2147483645,-2147483645,-2147483645,-2147483644,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483641,-2147483640,-2147483640,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483641,-2147483641,-2147483640,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483641,-2147483641,-2147483641,-2147483640,-2147483640,-2147483639,-2147483638,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483637,-2147483636,-2147483636,-2147483636]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483643,-2147483642,-2147483642,-2147483642,-2147483641,-2147483641,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483639,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483642,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641]
[0,-2147483647,-2147483646,-2147483645,-2147483645,-2147483645,-2147483644,-2147483644,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483643,-2147483643,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642]
[0,-2147483647,-2147483646,-2147483645,-2147483645,-2147483644,-2147483644,-2147483644,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641]
[0,-2147483647,-2147483647,-2147483646,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641]
[0,-2147483647,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483643,-2147483642,-2147483641,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483642,-2147483641,-2147483641,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640,-2147483640]
[0,-2147483647,-2147483646,-2147483645,-2147483645,-2147483645,-2147483645,-2147483645,-2147483645,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644]
[0,-2147483647,-2147483646,-2147483646,-2147483645,-2147483645,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641,-2147483641]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483641,-2147483640,-2147483640,-2147483640,-2147483640,-2147483639,-2147483639,-2147483638,-2147483637,-2147483636,-2147483636,-2147483636,-2147483635,-2147483635,-2147483635,-2147483635,-2147483635,-2147483635]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483643,-2147483642,-2147483642,-2147483642,-2147483641,-2147483641,-2147483641,-2147483641,-2147483640,-2147483640,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639,-2147483639]
[0,-2147483647,-2147483646,-2147483646,-2147483645,-2147483645,-2147483645,-2147483645,-2147483644,-2147483644,-2147483644,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643,-2147483643]
[0,-2147483647,-2147483646,-2147483645,-2147483645,-2147483644,-2147483644,-2147483643,-2147483643,-2147483643,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483644,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483641,-2147483641,-2147483640,-2147483640,-2147483639,-2147483639,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638,-2147483638]
[0,-2147483647,-2147483646,-2147483645,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483644,-2147483643,-2147483643,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642,-2147483642]

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Ben Goodrich]

You may be able to reproduce with clang++ without optimization. I think we have some undefined behavior here somewhere.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Ben Goodrich]

Okay, this is not going to be a widespread problem but we should fix Stan. In line 17 of

https://github.com/stan-dev/example-models/blob/master/bugs_examples/vol1/dogs/dogs.stan

it goes

xa[dog, trial] <- xa[dog, trial] + Y[dog, k];

but xa[dog,trial] is undefined at this point. It looks like sometimes the compiler uses 0, sometimes the smallest integer, etc. I thought we were initializing everything to NaN but I guess not.

Ben

[Daniel Lee]

I can confirm. The model is incorrect. It never initializes xa or xs to 0. We probably should initialize transformed data that are ints to 0.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Daniel Lee]

Ok, so we initialize in the generated C++ code. I'm not sure why this doesn't work under g++ 4.9:
  xs = std::vector<std::vector<int> >(Ndogs,std::vector<int>(Ntrials,int(0)));

[Ben Goodrich]

I'm not sure why it doesn't work, but shouldn't we initialize to NaN? Initializing to zero can result in valid code with the wrong data.

Ben

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Daniel Lee]

int types don't have nan.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Ben Goodrich]

Maybe the problem is with the next few lines where we fill with a double?

        double DUMMY_VAR__(std::numeric_limits<double>::quiet_NaN());
        (void) DUMMY_VAR__;  // suppress unused var warning


        // initialize transformed variables to avoid seg fault on val access
        stan::math::fill(y,DUMMY_VAR__);
        stan::math::fill(xa,DUMMY_VAR__);
        stan::math::fill(xs,DUMMY_VAR__);

On Wednesday, June 3, 2015 at 12:36:32 AM UTC-4, Daniel Lee wrote:

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+unsubscribe@googlegroups.com.

[Daniel Lee]

yeah, I just saw that too. I think that's what causes the problem.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.

To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.

[Daniel Lee]

I just submitted a bug report for this: https://github.com/stan-dev/stan/issues/1477

[Bob Carpenter]

I think this should be easy to fix on the generated code
side.

- Bob

[Joshua N Pritikin]

On Tue, Jun 02, 2015 at 07:49:35AM -0400, Ben Goodrich wrote:
> Attached is a copy with the Poisson thing fixed locally, but in general
> it is just a subrepo of
> [1]https://github.com/stan-dev/rstan/
> (which doesn't have the Poisson thing fixed yet because the patch
> hasn't been applied to Stan yet).

OK, it was easy to fix our code. The only complaint I have is that issue
1383 is not addressed,

https://github.com/stan-dev/stan/issues/1383

--
Joshua N. Pritikin
Department of Psychology
University of Virginia
485 McCormick Rd, Gilmer Hall Room 102
Charlottesville, VA 22904
http://people.virginia.edu/~jnp3bc

[Bob Carpenter]

Sorry about that --- we never prioritized that issue, so it
wasn't showing up on our to-do lists in the issue tracker for
the release. I just labeled it 2.7.0++, so it should go into
the next release after the one we're working on now.

Or maybe now that we have builds that work for R, we can go
a bit more slowly on 2.7.0 and fix a bunch of the other 2.7.0++
issues.

- Bob

[Bob Carpenter]

Also, feel free to jump in and make a pull request doing it
the way you like. This is going to change when we pull the math
lib out, because it'll get its own version numbering.

- Bob

> On Jun 3, 2015, at 4:12 AM, Joshua N Pritikin <jpri...@pobox.com> wrote:
>