R help - Jul 2008 - R crash with ATLAS precompiled Rblas.dll on Windows XP Core2 Duo

I noticed a problem using R 2.7.1 on Windows XP SP2 with the precompiled
Atlas Rblas.dll.  Running the code below causes R to crash.  I started R
using Rgui --vanilla and am using the precompiled Atlas Rblas.dll from
cran.fhcrc.org dated 17-Jul-2007 05:04 for Core2 Duo.

The code that causes the crash:

x <- rnorm(100)
y <- rnorm(100)
z <- rnorm(100)
loess(z ~ x * y)

loess(z ~ x) does not cause a crash using the Atlas BLAS and neither does
running the above code with the Rblas.dll that came with R 2.7.1.  In
addition, the code runs fine using the Atlas BLAS under R 2.6.2.

The windows error information that is printed to the screen when R closes:

AppName: rgui.exe	 AppVer: 2.71.45970.0	 ModName: rblas.dll
ModVer: 2.51.42199.0	 Offset: 000501cc


sessionInfo returns:

R version 2.7.1 (2008-06-23) 
i386-pc-mingw32 

locale:
LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
States.1252;LC_MONETARY=English_United
States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

I checked the R FAQ, R for Windows FAQ, and the README associated with the
Atlas BLAS on CRAN and couldn't find any information related to possible
crash causes.  I've used the ATLAS BLAS for about 6 months on this machine
(it's a new machine) with R 2.6.2.

Using debug(stats:::simpleLoess), I've found that the crash occurs on the
first iteration of the line:

z <- .C(R_loess_raw, as.double(y), as.double(x), 
                as.double(weights), as.double(robust), as.integer(D), 
                as.integer(N), as.double(span), as.integer(degree), 
                as.integer(nonparametric), as.integer(order.drop.sqr), 
                as.integer(sum.drop.sqr), as.double(span * cell), 
                as.character(surf.stat), fitted.values = double(N), 
                parameter = integer(7), a = integer(max.kd), 
                xi = double(max.kd), vert = double(2 * D), vval = double((D
+ 
                  1) * max.kd), diagonal = double(N), trL = double(1), 
                delta1 = double(1), delta2 = double(1), as.integer(surf.stat
== 
                  "interpolate/exact"))

After that, I'm kind of stuck in terms of tracking it down.

Thanks for any input,

Jason Law
City of Portland, OR

Yes, that Rblas.dll is known to be faulty, and the person who built it is 
unable to re-build it.  It needs to be removed from CRAN.

(I've also tried to build on Core 2 Duo, and my Cygwin installation has a 
compiler crash during the build.)

On Tue, 8 Jul 2008, Law, Jason wrote:
> I noticed a problem using R 2.7.1 on Windows XP SP2 with the precompiled
> Atlas Rblas.dll.  Running the code below causes R to crash.  I started R
> using Rgui --vanilla and am using the precompiled Atlas Rblas.dll from
> cran.fhcrc.org dated 17-Jul-2007 05:04 for Core2 Duo.
>
> The code that causes the crash:
>
> x <- rnorm(100)
> y <- rnorm(100)
> z <- rnorm(100)
> loess(z ~ x * y)
>
> loess(z ~ x) does not cause a crash using the Atlas BLAS and neither does
> running the above code with the Rblas.dll that came with R 2.7.1.  In
> addition, the code runs fine using the Atlas BLAS under R 2.6.2.
>
> The windows error information that is printed to the screen when R closes:
>
> AppName: rgui.exe	 AppVer: 2.71.45970.0	 ModName: rblas.dll
> ModVer: 2.51.42199.0	 Offset: 000501cc
>
>
> sessionInfo returns:
>
> R version 2.7.1 (2008-06-23)
> i386-pc-mingw32
>
> locale:
> LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
> States.1252;LC_MONETARY=English_United
> States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252
>
> attached base packages:
> [1] stats     graphics  grDevices utils     datasets  methods   base
>
> I checked the R FAQ, R for Windows FAQ, and the README associated with the
> Atlas BLAS on CRAN and couldn't find any information related to
possible
> crash causes.  I've used the ATLAS BLAS for about 6 months on this
machine
> (it's a new machine) with R 2.6.2.
>
> Using debug(stats:::simpleLoess), I've found that the crash occurs on
the
> first iteration of the line:
>
> z <- .C(R_loess_raw, as.double(y), as.double(x),
>                as.double(weights), as.double(robust), as.integer(D),
>                as.integer(N), as.double(span), as.integer(degree),
>                as.integer(nonparametric), as.integer(order.drop.sqr),
>                as.integer(sum.drop.sqr), as.double(span * cell),
>                as.character(surf.stat), fitted.values = double(N),
>                parameter = integer(7), a = integer(max.kd),
>                xi = double(max.kd), vert = double(2 * D), vval = double((D
> +
>                  1) * max.kd), diagonal = double(N), trL = double(1),
>                delta1 = double(1), delta2 = double(1), as.integer(surf.stat
> =>                  "interpolate/exact"))
>
> After that, I'm kind of stuck in terms of tracking it down.
>
> Thanks for any input,
>
> Jason Law
> City of Portland, OR
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595