[R] Cannot update some packages after upgrade to 2.1.1

[Prof Brian Ripley]

On Wed, 20 Jul 2005, Uwe Ligges wrote:

> Prof Brian Ripley wrote:
>
>> On Wed, 20 Jul 2005, Kevin E. Thorpe wrote:
>> 
>> 
>>> Thank you for the information. I have contacted the RPM maintainer and am 
>>> awaiting a response.
>>> 
>>> It occurs to me that my "problem" could also be fixed by putting ATLAS on 
>>> my system. Are there advantages to doing that or any reasons not to?
>> 
>> 
>> It is a good idea unless you share R systems between different 
>> architectures (even down to the vintage of P4 or Xeon chips). We do and so 
>> tend to avoid ATLAS (which gets statically compiled in by default).
>
>
> We are using you ATLAS-optimized Rblas.dll files for our facutly wide Windows 
> installation.
>
> Additional to the regular .../bin directory, we have copies called 
> .../bin-At, .../bin-P3, .../bin-P4 but the regular Rblas.dll replaced by 
> those optimized Rblas.dll files.
> Hence one can simply call .../bin-P4/RGui.exe on any P4/Xeon machine and do 
> not need a seperate installation for another Athlon machine using the 
> .../bin-At/RGui.exe, for example.

That is fine on Windows, which has a separate Rblas.dll exactly to enable 
this to be done.

You can also do it on Debian Linux, which has a dynamic libatlas (as I 
understand it).

On other systems with a static ATLAS library, the BLAS is compiled into 
the R installation in many (up to hundreds) of places, including each 
package that uses a BLAS.   It's all too easy to build R on, say, an 
Athlon and discover it segfaults on a Xeon because an Athlon-optimized 
ATLAS has been compiled in. (I've done it more than once, and it needs the 
whole setup to be reinstalled, including many of the packages without it 
being obvious which ones).

>
>
> Uwe
>
>
>> 
>>> Prof Brian Ripley wrote:
>>> 
>>> 
>>>> -lf77blas is part of ATLAS, so I do suspect the RPM builder had ATLAS 
>>>> installed.
>>>> 
>>>> lme4 needs a compatible Matrix installed.
>>>> 
>>>> I do think installing from the sources would solve this, but probably you 
>>>> need to discuss this with the RPM maintainer as a dependency appears to 
>>>> be missing.
>>>> 
>>>> On Fri, 15 Jul 2005, Kevin E. Thorpe wrote:
>>>> 
>>>> 
>>>>> I just upgraded to version 2.1.1 (from 2.0.1) today.
>>>>> 
>>>>> 
>>>>>> R.version
>>>>> 
>>>>> _
>>>>> platform i686-pc-linux-gnu
>>>>> arch i686
>>>>> os linux-gnu
>>>>> system i686, linux-gnu
>>>>> status
>>>>> major 2
>>>>> minor 1.1
>>>>> year 2005
>>>>> month 06
>>>>> day 20
>>>>> language R
>>>>> 
>>>>> I am using SuSE 9.2 and did the upgrade using rpm -U with the RPM
>>>>> available on CRAN. After upgrading r-base, I ran update.packages().
>>>>> Four previously installed packages failed to update:
>>>>> 
>>>>> Matrix (0.95-5 to 0.97-4)
>>>>> gam (0.93 to 0.94)
>>>>> lme4 (0.95-3 to 0.96-1)
>>>>> mgcv (1.3-1 to 1.3-4)
>>>>> 
>>>>> In the case of Matrix, gam and mgcv I get the message:
>>>>> 
>>>>> [long path]/ld: cannot find -lf77blas
>>>>> 
>>>>> In the case of lme4 the messages are:
>>>>> 
>>>>> ** preparing package for lazy loading
>>>>> Error in setMethod("coef", signature(object = "lmList"), 
>>>>> function(object, :
>>>>> no existing definition for function 'coef'
>>>>> Error: unable to load R code in package 'lme4'
>>>>> Execution halted
>>>>> ERROR: lazy loading failed for package 'lme4'
>>>>> 
>>>>> I have searched the SuSE repository for any package that provides
>>>>> f77blas but came up empty.
>>>>> 
>>>>> I also could not identify any relevant messages in the mailing list
>>>>> archives.
>>>>> 
>>>>> Is R looking for that library because it was present on the machine
>>>>> the RPM was built on? Would building R myself solve the missing library
>>>>> problem or did I do something wrong?
>>>>> 
>>> 
>>> 
>>> -- 
>>> Kevin E. Thorpe
>>> Biostatistician/Trialist, Knowledge Translation Program
>>> Assistant Professor, Department of Public Health Sciences
>>> Faculty of Medicine, University of Toronto
>>> email: kevin.thorpe at utoronto.ca  Tel: 416.946.8081  Fax: 416.971.2462
>>> 
>>> 
>> 
>> 
>
>

-- 
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