[Rd] Possible bug in 2.11.x texmf makefile.in and some related things...

Marc Schwartz marc_schwartz at me.com
Mon Jul 5 14:24:38 CEST 2010


On Jul 5, 2010, at 1:50 AM, Prof Brian Ripley wrote:

> On Sun, 4 Jul 2010, Marc Schwartz wrote:
> 
>> Hi all,
>> 
>> A list of some possible issues:
>> 
>> 1. In R 2.11.x, in:
>> 
>> http://svn.r-project.org/R/branches/R-2-11-branch/share/texmf/
>> 
>> there are two files, jss.cls and jss.bst (for JSS), which appear to be new since 2.10.x. These files are not installed when building/installing R. It would appear that they are not included in:
>> 
>> https://svn.r-project.org/R/branches/R-2-11-branch/share/Makefile.in
>> 
>> The relevant code there is:
>> 
>> 	@for f in $(srcdir)/texmf/*.sty \
>> 		  $(srcdir)/texmf/*.fd; do \
>> 	  $(INSTALL_DATA) $${f} "$(DESTDIR)$(rsharedir)/texmf"; \
>> 	done
>> 
>> which would skip over the two jss files.
>> 
>> This issue came up when another useR (on F12) was building the zoo package with its vignette, which apparently uses jss.cls. Errors were of course observed. I thought that the error was limited to Fedora, but this is an issue in the source.
> 
> But there are a fair number of packages which need extra LaTeX files to build their vignettes, and e.g. the CRAN test systems have had to have these installed for a long time.  Certainly zoo has needed them from before 2.11.0, and many believe that a Open Source package needs to include copies of such files (see 'Writing R Extensions').


OK...so strictly speaking then, this is a bug in zoo(?), since it does not include those two files to build the relevant documents and would appear to presume that they are available on target build systems?  This is not an issue when installing the package, only when building it locally AFAICS, but lacking the two files, it would not otherwise pass R CMD check. The resultant PDF files are in the zoo source tarball and of course in the binaries.

What is interesting is that if one reviews the CRAN checks for zoo, there are notes for OSX related to jss.cls not being present. So I am guessing that the other OS CRAN build systems have the two files, but the OSX build system does not? That led to some confusion, as I was not sure whether the files were expected to be present, given that zoo passes the checks on the other OS's.


>> 2. On a related issue, the texmf tree appears to be modified in 2.12.x, with the latex and bibtex files being put into separate folders:
> 
> More precisely, it is a TDS-comformant tree now.


Makes sense....thanks.


> 
>> https://svn.r-project.org/R/trunk/share/texmf/
>> 
>> The makefile.in:
>> 
>> https://svn.r-project.org/R/trunk/share/Makefile.in
>> 
>> appears to be modified to handle that split:
>> 
>> 	@for f in $(srcdir)/texmf/bibtex/bst/*; do \
>> 	  $(INSTALL_DATA) $${f} "$(DESTDIR)$(rsharedir)/texmf/bibtex/bst"; \
>> 	done
>> 	@for f in $(srcdir)/texmf/tex/latex/*; do \
>> 	  $(INSTALL_DATA) $${f} "$(DESTDIR)$(rsharedir)/texmf/tex/latex"; \
>> 	done
>> 
>> 
>> So there was perhaps an oversight of sorts for 2.11.x in handling these two JSS related files.
> 
> That's a different hand.  I don't know if the files were intended to be installed in 2.11.x: they are in R-devel so eventually packages depending on R >= 2.12.0 will be able to rely on them.
> 
>> 3. A final note, which is that the NEWS file appears to be missing from R-Devel tonight:
>> 
>> https://svn.r-project.org/R/trunk/
>> 
>> I was trying to read it to note any comments relevant to the above.
> 
> Correct.  You need to look in https://svn.r-project.org/R/trunk/doc/NEWS.Rd
> which says
> 
>      \item File \file{NEWS} is now generated at installation with a
>      slightly different format: it will be in UTF-8 on platforms using
>      UTF-8, and otherwise in ASCII.
> 
>      \item \file{NEWS} is no longer in the sources, but generated as
>      part of the installation.  The primary source for changes is now
>      \file{doc/NEWS.Rd}.
> 
> A version of the NEWS file will be in the tarballs.


OK. Again, thanks for the clarifications.

Regards,

Marc



More information about the R-devel mailing list