Hello all,
After using Bioconductor with pre-built CDF evironments, I've finally
faced
a problem of generating a CDF environment for a custom Affy chip. The
makecdfenv utility worked just fine under Linux and the custom CDF
environment is already usable there. However, under Windows (and I do
need
this as well) the final step of installation (R CMD INSTALL
uwyrhodo1acdf)
from a command prompt keeps failing, regardless to presence of
installed
Perl (as recommened in the manual) etc., etc. The contents of the
folder
with the name of the new environment seem to be equal under both Linux
(after installation which is functional) and Windows (after incomplete
installation). Could someone help me with this issue? (I use the
latest
R/Bioconductor).
Thanks!
Oleg
-------------------------------------
Oleg Moskvin
Research Scientist
Dept of Molecular Biology
University of Wyoming
ovm at uwyo.edu
Hi Seth,
Thanks for the reply. The error message does not look too informative:
-----------------------------------------------------------------
C:\Program Files\R\R-2.3.0\bin>R CMD INSTALL uwyrhodo1acdf
*** Installation of uwyrhodo1acdf failed ***
Removing 'C:/PROGRA~1/R/R-23~1.0/library/uwyrhodo1acdf'
C:\Program Files\R\R-2.3.0\bin>
-----------------------------------------------------------------
As you see, this is run from the "bin" directory of R. I've copied
the
folder generated by makecdfenv there. This may look weird but running
it
from other places (such as /library) results in not finding the R
executable
at all! See below.
-----------------------------------------------------------------
C:\Program Files\R\R-2.3.0\bin>R CMD INSTALL uwyrhodo1acdf
*** Installation of uwyrhodo1acdf failed ***
Removing 'C:/PROGRA~1/R/R-23~1.0/library/uwyrhodo1acdf'
C:\Program Files\R\R-2.3.0\bin>
-----------------------------------------------------------------
This looks like the "path" problem, but frankly, I do not now how to
fix
this in Windows XP (years ago I did edit this under DOS ;)).
As another possibility, I could supply the original cdf file, if it
is
possible to generate the cdf-environment remotely and then install it
from
zip file. BTW, several other laboratories would be interested in it
(the
Rhodobacter genechip will be commercially available soon).
Thanks for all the help.
Oleg
> ----- Original Message -----
> From: "Seth Falcon" <sfalcon at="" fhcrc.org="">
> To: "Oleg Moskvin" <ovm at="" uwyo.edu="">
> Sent: Tuesday, May 30, 2006 4:51 PM
> Subject: Re: [BioC] CDF environment for a custom Affy chip
>
>
>> Hi Oleg,
>>
>> "Oleg Moskvin" <ovm at="" uwyo.edu=""> writes:
>>> After using Bioconductor with pre-built CDF evironments, I've
finally
>>> faced
>>> a problem of generating a CDF environment for a custom Affy chip.
The
>>> makecdfenv utility worked just fine under Linux and the custom CDF
>>> environment is already usable there. However, under Windows (and I
do
>>> need
>>> this as well) the final step of installation (R CMD INSTALL
>>> uwyrhodo1acdf)
>>> from a command prompt keeps failing, regardless to presence of
installed
>>> Perl (as recommened in the manual) etc., etc. The contents of the
folder
>>> with the name of the new environment seem to be equal under both
Linux
>>> (after installation which is functional) and Windows (after
incomplete
>>> installation). Could someone help me with this issue? (I use the
latest
>>> R/Bioconductor).
>>
>> Can you provide some further detail on the errors you get on
Windows?
>> Perhaps copy/paste the relevant part of the terminal session.
>>
>> Please try to give us as much detail as possible. For example,
exact
>> R version.
>>
>> + seth
>>
>
[I goofed my original response and didn't cc the list. With Oleg's
permission, I'm forwarding his response (with comments added)]
"Oleg Moskvin" <ovm at="" uwyo.edu=""> writes:
> Hi Seth,
>
> Thanks for the reply. The error message does not look too
informative:
>
>
> -----------------------------------------------------------------
> C:\Program Files\R\R-2.3.0\bin>R CMD INSTALL uwyrhodo1acdf
>
> *** Installation of uwyrhodo1acdf failed ***
>
> Removing 'C:/PROGRA~1/R/R-23~1.0/library/uwyrhodo1acdf'
>
> C:\Program Files\R\R-2.3.0\bin>
> -----------------------------------------------------------------
>
> As you see, this is run from the "bin" directory of R. I've copied
the
> folder generated by makecdfenv there. This may look weird but
running
> it from other places (such as /library) results in not finding the R
> executable at all! See below.
>
> -----------------------------------------------------------------
> C:\Program Files\R\R-2.3.0\bin>R CMD INSTALL uwyrhodo1acdf
>
> *** Installation of uwyrhodo1acdf failed ***
>
> Removing 'C:/PROGRA~1/R/R-23~1.0/library/uwyrhodo1acdf'
>
> C:\Program Files\R\R-2.3.0\bin>
> -----------------------------------------------------------------
I think you copy/pasted the wrong thing for the second example. But
your path issue is almost certainly the problem. The chain of tools
required to build packages from their source directories on Windows is
quite particular; if you don't get things just right, it won't work.
> This looks like the "path" problem, but frankly, I do not now how to
> fix this in Windows XP (years ago I did edit this under DOS ;)).
We have a document that describes building R from source on Windows.
It is here: http://wiki.fhcrc.org/bioc/HowTo/Build_R_on_Windows
You may not need a complete source build for what you want to do on
Windows, but you _will_ need a complete and working tool chain. The
link I've provided has details on this, including PATH stuff.
To set your PATH on Windows: right-click on My Computer, select
Properties, Advanced, Environment Variables (this is from memory, may
not be exactly right).
> As another possibility, I could supply the original cdf file, if it
is
> possible to generate the cdf-environment remotely and then install
it
> from zip file. BTW, several other laboratories would be interested
in
> it (the Rhodobacter genechip will be commercially available soon).
Seems like you are close to getting things working, but we could
arrange to build the package if you can't get it working.
And if the result would be of general interest to folks, we can also
consider adding it to our hosted repository.
Best,
--
+ seth
On 5/31/06, Seth Falcon <sfalcon at="" fhcrc.org=""> wrote:
> [I goofed my original response and didn't cc the list. With Oleg's
> permission, I'm forwarding his response (with comments added)]
>
>
> "Oleg Moskvin" <ovm at="" uwyo.edu=""> writes:
> > Hi Seth,
> >
> > Thanks for the reply. The error message does not look too
informative:
> >
> >
> > -----------------------------------------------------------------
> > C:\Program Files\R\R-2.3.0\bin>R CMD INSTALL uwyrhodo1acdf
> >
> > *** Installation of uwyrhodo1acdf failed ***
> >
> > Removing 'C:/PROGRA~1/R/R-23~1.0/library/uwyrhodo1acdf'
> >
> > C:\Program Files\R\R-2.3.0\bin>
> > -----------------------------------------------------------------
> >
> > As you see, this is run from the "bin" directory of R. I've copied
the
> > folder generated by makecdfenv there. This may look weird but
running
> > it from other places (such as /library) results in not finding the
R
> > executable at all! See below.
> >
> > -----------------------------------------------------------------
> > C:\Program Files\R\R-2.3.0\bin>R CMD INSTALL uwyrhodo1acdf
> >
> > *** Installation of uwyrhodo1acdf failed ***
> >
> > Removing 'C:/PROGRA~1/R/R-23~1.0/library/uwyrhodo1acdf'
> >
> > C:\Program Files\R\R-2.3.0\bin>
> > -----------------------------------------------------------------
>
> I think you copy/pasted the wrong thing for the second example. But
> your path issue is almost certainly the problem. The chain of tools
> required to build packages from their source directories on Windows
is
> quite particular; if you don't get things just right, it won't work.
>
> > This looks like the "path" problem, but frankly, I do not now how
to
> > fix this in Windows XP (years ago I did edit this under DOS ;)).
>
> We have a document that describes building R from source on Windows.
> It is here: http://wiki.fhcrc.org/bioc/HowTo/Build_R_on_Windows
>
> You may not need a complete source build for what you want to do on
> Windows, but you _will_ need a complete and working tool chain. The
> link I've provided has details on this, including PATH stuff.
>
> To set your PATH on Windows: right-click on My Computer, select
> Properties, Advanced, Environment Variables (this is from memory,
may
> not be exactly right).
I've got a fairly generic RCMDprompt.bat script, which opens a Windows
Command prompt with PATH etc setup. It assumes default installation
paths for many of the tools, and it figures out (from the Windows
Registry) where the R installation lives. Thus, if everything works
as planned you should be able to just double click the script and get
a working environment for building packages, given that you have
installed all tools that is (see script for instructions). Download:
http://www.maths.lth.se/help/R/RCMDprompt.bat
/Henrik
> > As another possibility, I could supply the original cdf file, if
it is
> > possible to generate the cdf-environment remotely and then install
it
> > from zip file. BTW, several other laboratories would be interested
in
> > it (the Rhodobacter genechip will be commercially available soon).
>
> Seems like you are close to getting things working, but we could
> arrange to build the package if you can't get it working.
>
> And if the result would be of general interest to folks, we can also
> consider adding it to our hosted repository.
>
> Best,
>
> --
> + seth
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives:
http://news.gmane.org/gmane.science.biology.informatics.conductor
>
>
