[pkg-bioc] State of BioC compilations

[Dirk Eddelbuettel]

Hi Steffen,

Thanks for all the work on this. Looks like you're moving along, with some
speed bumps on the way. Let's see if we can help with those.

On 9 February 2007 at 12:00, Steffen Moeller wrote:
| Hi,
| 
| in principle it all works. Some issues I get with my 2.4 R installation that 
| does not accept some apparent novel language construct 1L or 0L, which 
| apparently  denote a vector of zeros or ones.

Could you (briefly yet concisely) describe what happens, and where?  I am not
aware of unpatched faults in Debian's R 2.4.1, and R Core would not change
the language just for kicks. Or are you referring to C/C++ constructs?  Here
	long foo = 0L;
is legal AFAIK and no compiler should stumble.  
 
| I have ongoing problems to install the graph package. Both the devel and the 
| release versions build a nice packages but nothing pops up in the library() 
| call. Any clues? This blocks some much seeked after packages like Rgraphviz.

It probably means that dpkg/debhelper et al install in the wrong place, so
nothing gets tar'ed up. Doing the package manually with the generated
debian/rules and turning debhelper verbosity on may help.

| Something that does not work are the dependencies. This is real bad. And I 
| wished, a compilation would not be attempted when a known build-dep is not 
| installed. I had a quick look at the AptPkg perl package but only a quick 
| one. No proper dependencies and the continues broken build attempts are no 
| fun.

I do not understand. Dependencies are resolved by dpkg, we create them based
on control info we read. That already worked in Albrecht's scrip that I
hacked way when. So again: what fails, where, how?  Reproducible example?
| 
| The download script works but does not download any indirectly required 
| packages. Hence the initial constraint on the lite package is unpractical. We 
| should eventually come up with a concise list of packages that we want in 
| Debian and directly specify just that.

As I understand it, that is a BioC design choice. They make it hard to read
the directory, so we have to query via R code as BioCLite.R et al do.

Both CRAN and BioC also get too big, so we should have 'Debian task views'
(better term suggestions welcomes) that cover subareas.
 
| Does anybody have the r-omega-rcurl package available, even for submission to 
| Debian maybe? It is required from several non-nonsense BioC packages.

Was never packaged AFAIK but IIRC does exist in CRAN. Try building
r-cran-curl, maybe?

| Package count:  ls builds/r-bioc-* | wc -l
| 54

Not bad!  54 down, 900 at CRAN to go :)

Dirk

-- 
Hell, there are no rules here - we're trying to accomplish something. 
                                                  -- Thomas A. Edison