R help - Mar 2009 - R 2.8.1 and 2.9 alpha crash when running survest of Design package

Dear Prof Harrell and everyone,

 

My PC: Window XP service pack 3 and service pack 2

R version 2.8.1 and 2.9 alpha

 

For the last 3 days, after updating R, my two computers have been facing
problems when running existing and runable R commands that involves with
Design package

 

I attempt to use 'survest', but I failed all the times with R (both
2.8.1
and 2.9 alpha) being shut down immediately with following error report
messages.

AppName: rgui.exe AppVer: 2.90.48212.0   ModName: survival.dll

ModVer: 0.0.0.0   Offset: 00007749

 

However, if I run these commands on other computers which have not been
updated for 2 week, they run OK

 

Could you please consider the matter and give me advice

 

I am looking forward to hearing from you soon

 

Regards

Nguyen D Nguyen
Garvan Institute of Medical Research

Sydney, Australia


	[[alternative HTML version deleted]]

The Design package is incompatible with updates to the survival package (version
2.35 and higher) that were made for version 2.9.0. It calls some internal
fitting functions (coxreg.fit, agreg.fit) whose arguments have changed.

According to the CRAN checks, about a dozen other packages were also affected by
other changes in the update, but most just give an error message rather than
crash. Maintainers of all the packages have been notified and given my best
guess at the reason for their specific incompatibility (and an offer of further
assistance if necessary)

You may need to downgrade to version 2.34 of the survival package until Design
is updated.

      -thomas



On Fri, 27 Mar 2009, Nguyen Dinh Nguyen wrote:
> Dear Prof Harrell and everyone,
>
>
>
> My PC: Window XP service pack 3 and service pack 2
>
> R version 2.8.1 and 2.9 alpha
>
>
>
> For the last 3 days, after updating R, my two computers have been facing
> problems when running existing and runable R commands that involves with
> Design package
>
>
>
> I attempt to use 'survest', but I failed all the times with R (both
2.8.1
> and 2.9 alpha) being shut down immediately with following error report
> messages.
>
> AppName: rgui.exe AppVer: 2.90.48212.0   ModName: survival.dll
>
> ModVer: 0.0.0.0   Offset: 00007749
>
>
>
> However, if I run these commands on other computers which have not been
> updated for 2 week, they run OK
>
>
>
> Could you please consider the matter and give me advice
>
>
>
> I am looking forward to hearing from you soon
>
>
>
> Regards
>
> Nguyen D Nguyen
> Garvan Institute of Medical Research
>
> Sydney, Australia
>
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> 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.
>
Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle

A couple additions to Thomas's message.

  The 'survest' function in design directly called C routines in the
survival
package.  The argument list to the routines changed due to the addition of 
weights; calling a C routine with the wrong arguments is one of the more 
reliable ways to crash a program.  The simplest (short term) solution is to use 
survfit for your curves rather than survest.  Frank Harrell has been aware of 
the issue for several weeks and is working hard on solving it.  The simple fix 
is a few minutes, but he's thinking about how to avoid any future problems. 
The
C routines in survival change arguments VERY rarely, but direcly calling the 
routines of another package is considered dangerous in general.
  
  Most breakage was less severe.  For instance there were a couple of errors in 
the PBC data set.  I fixed these, and also replaced all the "999"
codes with NA
to make it easier to use.  Some other packages use this data.  (My name is on 
most of the PBC papers and I have the master PBC data with all labs, patient id,
etc, but I was not the source of the first data set).  
  
  We'll be keeping an eye on the R list as the package rolls out; sending a 
message directly to Thomas and/or I would also be appreciated for issues like 
this.
  
  	Terry Therneau