[Bioc-devel] EBImage: Devel version on Windows not building

[Patrick Aboyoun]

Greg and Wolfgang,
The Windows build machines all have GTK+ and ImageMagick installed on 
them so you can link to the appropriate dlls during the build process. I 
just took a cursory look at the EBImage package, and it looks like the 
change would be minor.

1) Remove the pre-built dll and all the configuration code that manages it.
2) Move the GTK_BASEPATH and IM_BASEPATH definitions out of the 
Makevars.win file, make them Windows Environment Variables, and check to 
ensure they are defined during the configuration process.

I just added these two environment variables to the windows build 
machines so they are ready to go. If we run into other snags, we can 
make additional tweaks.


Patrick



Wolfgang Huber wrote:
> Hi Patrick et al. in Seattle
>
> maybe the main point of Greg's message got lost, so here it is again:
> would it be possible to compile the C code in EBImage from source on the
> Seattle build system?
>
> This would seem to be preferable to the current solution, in which a
> .dll file that was made by Oleg is used, which is checked into svn:
> https://hedgehog.fhcrc.org/bioconductor/trunk/madman/Rpacks/EBImage/inst/winlibs/EBImage.dll
>
> ------------------------------------------------------------------------
> r21716 | osklyar at ebi.ac.uk | 2006-12-30 23:02:15 +0000 (sab, 30 dic
> 2006) | 1 line
>
> added precompiled windows dll to save work of Bioc maintainers in
> installing and maintaining GTK+ and ImageMagick on Windows build
> machines! I  will recompile the windows dll myself, the target dll is
> only 38K large, so this is not a big problem. The method seems to be
> consistent with R CMD check on Linux. The idea is: inst/winlibs contains
> the precompiled dll, which is moved to inst/libs on build and then
> copied into libs by the installer; if dll is found in winlibs then the
> src directory is removed to prevent compilation mistakes, if not -- the
> package is recompiled. R CMD check does not check winlibs and does not
> warn (as it does if the dll is put into libs) so the only drawback is
> adding a unnecess ary windll to all *nix installations.
>
>
> Best wishes
>  Wolfgang
>
> ------------------------------------------------------------------
> Wolfgang Huber  EBI/EMBL  Cambridge UK  http://www.ebi.ac.uk/huber
>
>
> 13/08/2008 14:00 Gregoire Pau scripsit
>   
>> Hi,
>>
>> I'm the new maintainer of EBImage.
>>
>> The DLL EBImage.dll has been manually compiled some time ago and is not
>> in sync with the current version. It would be really nice that the
>> Windows version can be automatically built by the automatic Bioconductor
>> building framework, but EBImage relies on two external libraries:
>> GTK+2.0 and ImageMagick and I'm not sure whether they are available on
>> the Bioconductor building machines.
>>
>> What is the general policy for packages that rely on external libraries
>> ? Can I rely on some pre-installed libraries or should I build them
>> manually for some platforms ?
>>
>> Thanks !
>> Best regards,
>>
>> Greg
>>
>> Henrik Bengtsson wrote:
>>     
>>> Hi,
>>>
>>> thanks for fixing the DESCRIPTION file for EBImage.  FYI, the Windows
>>> version does not build:
>>>
>>> * creating vignettes ... ERROR
>>> Error in inDL(x, as.logical(local), as.logical(now), ...) :
>>>   unable to load shared library
>>> 'D:/biocbld/bbs-2.3-bioc/tmpdir/Rinst415629052/EBImage/libs/EBImage.dll':
>>>   LoadLibrary failure:  The specified procedure could not be found.
>>>
>>> Error: processing vignette 'AnalysisWithEBImage.Rnw' failed with
>>> diagnostics:
>>>  chunk 1 (label=loadlib)
>>> Error : package/namespace load failed for 'EBImage'
>>> Execution halted
>>>
>>> More:
>>> http://bioconductor.org/checkResults/2.3/bioc-LATEST/EBImage/gewurz-buildsrc.html
>>>
>>>
>>> I noticed this already a few weeks ago (before the typo in DESCRIPTION
>>> was added) and hoped it would solve "itself", but it is still there.
>>> Not sure if it is an EBImage or a BioC server config problem.
>>>
>>> Cheers
>>>
>>> Henrik
>>>
>>> PS. I can solve all these problems for myself, but I reporting it for
>>> the sake of automatic installation etc. DS.
>>>
>>> _______________________________________________
>>> Bioc-devel at stat.math.ethz.ch mailing list
>>> https://stat.ethz.ch/mailman/listinfo/bioc-devel
>>>