[R-SIG-Mac] [BioC] affy won't load

[Patrick Aboyoun]

Thanks to all for bringing this issue of the BioC 2.2 Max OS X binary 
packages to light, and I apologize for any inconvenience these 
non-standard binary packages have caused the BioC user community. If all 
goes well with the build system tomorrow morning, the Mac OS X binary 
packages will be produced under the standard Mac OS X R framework. I 
will send out an e-mail tomorrow morning notifying you on its status.

I don't want to get too much into the gory details of the build system, 
but I will provide you with a high level synopsis on what caused the 
breakdown. Back when the Bioconductor core team released BioC 2.2 in 
May, we had a last minute scramble for the Mac OS X binary packages 
builds.  Up until release time, we were building the packages on Mac OS 
X Leopard (10.5.2). It was brought to our attention the week of the 
release that those builds were not working on Mac OS X Tiger. When we 
switched to a Tiger build machine, we experienced problems with the png 
device (type = "quartz", the default, failed to work in our build 
environment). In order to be expedient we used a custom build of R 2.7.0 
to get around this issue. This had the unintended consequence that you 
discovered. We have since reverted back to using the standard R 2.7.1 
build for the Mac and are using a configuration file to set 
options(bitmapType = "cairo") to get around the png device issue. These 
new Mac binary package builds will be linking into the dynamic libraries 
located in the standard R directory 
/Library/Frameworks/R.framework/Resources/lib.


Cheers,
Patrick



Simon Urbanek wrote:
>
> On Jun 19, 2008, at 2:03 PM, Kasper Daniel Hansen wrote:
>
>> Another way Jason can get up and run quickly is to do something like
>>
>> R> options(pkgType = "source")
>>
>> and then do biocLite. That is probably a better solution than trying 
>> to install gettext
>>
>
> Definitely, and especially since it doesn't help (see my comments to 
> Steve below).
>
> I wouldn't recommend installing gettext (unless you know what you're 
> doing) simply because it will modify your system such that all future 
> configurations will detect it and behave differently. As long as you 
> don't plan to copy anything to another machine it is not a big deal, 
> but if you do, you'll learn very quickly why that's a bad idea ;). 
> Nonetheless, there are valid cases.
>
> However, as Jan pointed out correctly, this is not the only problem 
> that BioC builds have, unfortunately. I can only repeat my plea that I 
> sent while ago to the BioC team to fix their Mac builds.
>
>
>> Kasper
>>
>> On Jun 19, 2008, at 10:18 AM, Steve Lianoglou wrote:
>>
>>> Hi,
>>>
>>> On Jun 19, 2008, at 12:58 PM, Kasper Daniel Hansen wrote:
>>>
>>>> Ok, I will take this over to the R-SIG-Mac list and cc Herve who is 
>>>> the build manager of Bioconductor. I suspect that something is 
>>>> wrong with the Bioconductor builds.
>>>>
>>>> Synopsis: Jason is running R-2.7.0 from CRAN on an Intel macbook 
>>>> running 10.5.3. He says he has Xcode 3.0 installed. He tries to 
>>>> install the binary version of aff from Bioconductor using the normal
>>>> biocLite("affy")
>>>> However, even though the tarball downloads and gets installed, the 
>>>> package won't load. It complains that it needs the library 
>>>> /usr/local/lib/libintl.3.dylib (and an otool -L really confirms 
>>>> this), but apparently this files does not exists. I am a bit 
>>>> surprised that the binary version links to something in /usr/local. 
>>>> I have this file on my PPC G4 running 10.5.3. As I see it, this 
>>>> should work out of the box with the system Jason is running. Does 
>>>> anyone have any good ideas? My hunch is that something might be up 
>>>> with the binary builds?
>>>
>>> Just so Jason can get up and running quickly (and not have to wait 
>>> for this to get sorted out in the builds or whatever), I would 
>>> recommend as a band-aid procedure to just install gettext for now, 
>>> as I am fairly certain that this will provide the 
>>> /usr/local/lib/libintl.3.dylib that he is missing.
>>>
>
> Are you? Did you actually test it? I'm pretty sure that you're wrong, 
> because the commands below compile libintl.8.dylib(!) and thus won't 
> help (that one is fine and part of your Gtk binary). I'm still puzzled 
> at where the BioC team gets libintl.3 ...
>
> Cheers,
> Simon
>
>
>>> I gather that he might not be very comfortable from the command 
>>> line, so following these steps in a terminal session should do the 
>>> trick:
>>>
>>> cd /tmp
>>> curl http://mirrors.usc.edu/pub/gnu/gettext/gettext-0.17.tar.gz | 
>>> tar xz
>>> cd gettext-0.17
>>> ./configure
>>> make
>>> sudo make install
>>>
>>> HTH,
>>> -steve
>>>
>>> -- 
>>> Steve Lianoglou
>>> Graduate Student: Physiology, Biophysics, and Systems Biology
>>> Weill Medical College, Cornell University
>>>
>>> http://cbio.mskcc.org/~lianos
>>
>> _______________________________________________
>> R-SIG-Mac mailing list
>> R-SIG-Mac at stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>>
>>
>
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-mac