[Bioc-devel] Some build system news

Kasper Daniel Hansen khansen at stat.Berkeley.EDU
Tue Dec 19 00:30:01 CET 2006


On Dec 18, 2006, at 7:55 PM, Herve Pages wrote:

> Kasper Daniel Hansen wrote:
>> On Dec 15, 2006, at 4:28 AM, Herve Pages wrote:
>>
>>> "Main" build machine means that a devel package is made publicly
>>> available
>>> (by being pushed to http://bioconductor.org/packages/2.0/bioc/) only
>>> if it
>>> passes the CHECK on 'lamb1' with no error (it can have warnings  
>>> though).
>>
>> I assume you have thought this through. I just want to point out  
>> the the
>> lamb1 system is one specific architecture and there may be  
>> (development)
>> packages that have problems only for specific architectures. This  
>> is of
>> course mostly relevant for non-R code, and I may be a bit biased  
>> since I
>> have worked on two packages where the platform differences really  
>> mattered.
>>
>
> Hi Kasper,
>
> I should have said:
>
>   "Main" build machine means that the _src_ tarball (.tar.gz) of a  
> package is
>    made publicly available if it passes the CHECK on 'lamb1' with  
> no error.
>
> For binary packages things are a little bit different:
>   - lemming is our "main" Windows build machine
>   - We don't build Mac binaries: we only build and check the source  
> packages
>     on derby, our Mac build machine. Universal binaries are built  
> externally by
>     Simon Urbanek (results are here http://r.research.att.com/ 
> reports/) but for
>     release packages only.
>
> I agree that the lamb1 system is still one specific platform among
> many other Unix/Linux platforms. The underlying question is: on how  
> many
> platforms should a source package build and pass check before it  
> can be
> considered good enough to be made publicly available? And what is  
> the exact
> definition of "platform" to start with: is it just the  
> microprossesor type
> or the n-uplet (microprossesor type, OS type, OS version,  
> compilers, etc...)?

Now I get it. I agree with Seth that there is no compelling reason to  
change it. For me, the interesting part is that check gets run on all  
platforms irrespectively of the result on one of them. Whether or not  
the tarball is released is less important.

However, if I were to pick one core platform it would be 32 bit Linux.

Kasper


> Best,
> H.
>
>
>> Kasper
>>
>>> Also, maybe you've been wondering what's going on with some package
>>> description
>>> pages:
>>>
>>>   http://bioconductor.org/packages/2.0/bioc/
>>>
>>> As you can see, the individual pages for the packages listed  
>>> below the
>>> GGtools
>>> package are missing. This problem is related to the migration of  
>>> all our
>>> scripts from gopher5 to lamb1 and should get back to a normal state
>>> tomorrow
>>> around noon.
>>>
>>> Also the build results for our new Windows server (liverpool) are
>>> not looking good for an obscure pdflatex.exe error that occurs
>>> when building most of the vignettes. We'll investigate this and
>>> try to fix ASAP. In the meantime, if someone has already seen this
>>> problem
>>> before and can provide some usefull feedback, s/he is more than  
>>> welcome.
>>> A typical example can be seen here:
>>>
>>>
>>> http://bioconductor.org/checkResults/2.0/liverpool-ABarray- 
>>> buildsrc.html
>>>
>>> Thanks!
>>>
>>> Best,
>>> H.
>>>
>>> _______________________________________________
>>> Bioc-devel at stat.math.ethz.ch mailing list
>>> https://stat.ethz.ch/mailman/listinfo/bioc-devel
>>



More information about the Bioc-devel mailing list