Here are a few comments on the new release (1.2.0), most about R CMD check. I have managed to install and check my libraries only making one change to a C function (see point 1) and one to an R function because of the way formulae now look for their data. That is the good news. Of course, more extensive checks are liable to find further problems. 1. I need to be able to define F77_SYMBOL. This now in Defn.h which is not accessible to us mortals. Those few lines need to be moved into one of the accessible header files. 3. Makefiles in pkg/src are no longer read properly. This creates a big problem for me in library development. 3. DESCRIPTION has to have licence mis-spelled as license for the check to work. 4. check produces unasked for tex output (that I would like to be able to shut off) then an error that linktocpage is undefined. 5. check always produces an error message at the end: Attempt to free unreferenced scalar during global destruction. I am not sure if this is a reference to Chernobyl but I do not understand it. Other requests about the library check: 1. I would very much prefer that the check directory remain within the library directory instead of parallel as pkg.Rcheck to avoid proliferation of directories at that level. 2. Why can't check detect that the help pages are up-to-date (i.e. from a previous check) instead of redoing them every time? Jim -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
>>>>> On Fri, 15 Dec 2000 16:12:06 +0100 (MET), >>>>> james lindsey (jl) wrote:> Here are a few comments on the new release (1.2.0), most about R CMD > check. I have managed to install and check my libraries only making one > change to a C function (see point 1) and one to an R function because > of the way formulae now look for their data. That is the good news. Of > course, more extensive checks are liable to find further problems. > 1. I need to be able to define F77_SYMBOL. This now in Defn.h which is > not accessible to us mortals. Those few lines need to be moved into > one of the accessible header files. > 3. Makefiles in pkg/src are no longer read properly. This creates a > big problem for me in library development. what do you mean by ``not read properly''? if a makefile exists it should be used, actually `R CMD check' only uses the Makefile during a call to `R CMD INSTALL', so do you mean that INSTALL is broken (which would be bad news indeed)? > 3. DESCRIPTION has to have licence mis-spelled as license for the > check to work. hmm, that field had that name (with that spelling) for quite some time now. I think things get complicated if we allow for different spellings of variables for different observations (I basically build a dataframe from all DESCRIPTION files) ... but I'll leave that for native English speakers to decide. > 4. check produces unasked for tex output (that I would like to be able > to shut off) then an error that linktocpage is undefined. option --no-latex supresses the latex part. > 5. check always produces an error message at the end: > Attempt to free unreferenced scalar during global destruction. > I am not sure if this is a reference to Chernobyl but I do not > understand it. I get no error message: which version pf perl on which platform are you using? > Other requests about the library check: > 1. I would very much prefer that the check directory remain within the > library directory instead of parallel as pkg.Rcheck to avoid > proliferation of directories at that level. use option --outdir to change the default > 2. Why can't check detect that the help pages are up-to-date > (i.e. from a previous check) instead of redoing them every time? use option --no-clean for not removing the old files. the defaults are set to do ALL check FROM SCRATCH because that is the safest thing to do (meaning you can trust the results and don't miss any checks offered), but you can turn off most of the checks using command line options. hope this helps, fritz -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
On Fri, 15 Dec 2000 james.lindsey@luc.ac.be wrote:> Here are a few comments on the new release (1.2.0), most about R CMD > check. I have managed to install and check my libraries only making one > change to a C function (see point 1) and one to an R function because > of the way formulae now look for their data. That is the good news. Of > course, more extensive checks are liable to find further problems. > > 1. I need to be able to define F77_SYMBOL. This now in Defn.h which is > not accessible to us mortals. Those few lines need to be moved into > one of the accessible header files.You could just use the calls that are documented, in R-exts, that is F77_CALL and F77_NAME! F77_SYMBOL has not been documented as a user entry point since at least 1.0.0, and has never been part of the API. Only one of the 82 packages on CRAN was using it.> 3. Makefiles in pkg/src are no longer read properly. This creates a > big problem for me in library development.Having Makefiles in packages causes users a lot of problems, but as far as I know they still work (and so still break things as we found with polymars last week). Again, the preferred mechanisms have been in R-exts for a year or so.> 3. DESCRIPTION has to have licence mis-spelled as license for the > check to work.Actually, as `License:'> 4. check produces unasked for tex output (that I would like to be able > to shut off) then an error that linktocpage is undefined.It was asked for as it is a default option. Looks like your tex setup is broken.> 5. check always produces an error message at the end: > Attempt to free unreferenced scalar during global destruction. > I am not sure if this is a reference to Chernobyl but I do not > understand it.Is your Perl up to date?> Other requests about the library check: > 1. I would very much prefer that the check directory remain within the > library directory instead of parallel as pkg.Rcheck to avoid > proliferation of directories at that level. > > 2. Why can't check detect that the help pages are up-to-date > (i.e. from a previous check) instead of redoing them every time?By default, it does a clean install. You can use --no-clean, documented as one of the options. -- Brian D. Ripley, ripley@stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272860 (secr) Oxford OX1 3TG, UK Fax: +44 1865 272595 -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-devel-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._