R devel - Jan 2006 - Problem installing from source: no CONTENTS files

I've just had the error below while trying to install a package from
source under R2.2.1 and Windows XP. I recall encountering this
sporadically in the past. It is a pretty confusing message and took me
quite some time to figure out; I've seen queries about it on the R site
search, but couldn't find a reply. There's an attempted diagnosis and
suggested cure below.

D:\r2.0\debug>rcmd INSTALL --library=d:/rpackages/r2.2.1 --docs=normal
debug

---------- Making package debug ------------
<<snipped>>
installing indices
cat: C:/R/RW2021/library/*/CONTENTS: No such file or directory

What I think is happening is this. Because I installed R from binary
*without* HTML help, no CONTENTS files are created for the packages.
When running RCMD INSTALL, this causes an error in the MakePkg file (in
src/gnuwin32 on my machine), in the "indices" section at the line
starting @$(CAT) below. Commenting out the CAT line seems to work fine,
in that installation proceeds OK, and regular non-HTML help and
help.search are available for the package. The comment two lines up
suggests that the behaviour is a bug; maybe a TRY-style wrapper could be
used? If not, a different error message would be handy. I don't
understand enough about makefiles to suggest a patch, unfortunately. 

indices:
	@$(ECHO) "  installing indices"
	@$(ECHO) "invisible(.libPaths(c(.Library,\"$(RLIB)\",
.libPaths()))); tools:::.install_package_indices('.',
'"$(DPKG)"')" | \
	  R_DEFAULT_PACKAGES=NULL LC_COLLATE=C R_OSTYPE=windows
$(REXE)
> /dev/null
# need not have HTML installed when building packages
	@$(MKDIR) -p $(RHOME)/doc/html/search
	@$(CAT) $(RHOME)/library/*/CONTENTS >
$(RHOME)/doc/html/search/index.txt


Mark Bravington
CSIRO Mathematical & Information Sciences
Marine Laboratory
Castray Esplanade
Hobart 7001
TAS

ph (+61) 3 6232 5118
fax (+61) 3 6232 5012
mob (+61) 438 315 623

This will only happen if you installed R without HTML help and then 
install a package *with* HTML help *and* do so into a non-standard 
directory.  That seems an arcane thing to do: why do you want HTML help 
for some packages and not others?  You will end up with a partially 
functional system.

The fix is simple: put '-' at the beginning of that line (after the tab)
in MakePkg.  I've put a more complete solution in R-devel.

On Mon, 9 Jan 2006 Mark.Bravington at csiro.au wrote:
> I've just had the error below while trying to install a package from
> source under R2.2.1 and Windows XP. I recall encountering this
> sporadically in the past. It is a pretty confusing message and took me
> quite some time to figure out; I've seen queries about it on the R site
> search, but couldn't find a reply.
I don't believe those are the same thing.
> There's an attempted diagnosis and
> suggested cure below.
>
> D:\r2.0\debug>rcmd INSTALL --library=d:/rpackages/r2.2.1 --docs=normal
> debug
>
> ---------- Making package debug ------------
> <<snipped>>
> installing indices
> cat: C:/R/RW2021/library/*/CONTENTS: No such file or directory
>
> What I think is happening is this. Because I installed R from binary
> *without* HTML help, no CONTENTS files are created for the packages.
> When running RCMD INSTALL, this causes an error in the MakePkg file (in
> src/gnuwin32 on my machine), in the "indices" section at the line
> starting @$(CAT) below. Commenting out the CAT line seems to work fine,
> in that installation proceeds OK, and regular non-HTML help and
> help.search are available for the package. The comment two lines up
> suggests that the behaviour is a bug; maybe a TRY-style wrapper could be
> used? If not, a different error message would be handy.
The error message is completely informative: why would you want to change 
it?
> I don't understand enough about makefiles to suggest a patch, 
> unfortunately.
A good learning project for you?

>
> indices:
> 	@$(ECHO) "  installing indices"
> 	@$(ECHO) "invisible(.libPaths(c(.Library,\"$(RLIB)\",
> .libPaths()))); tools:::.install_package_indices('.',
'"$(DPKG)"')" | \
> 	  R_DEFAULT_PACKAGES=NULL LC_COLLATE=C R_OSTYPE=windows $(REXE)
>> /dev/null
> # need not have HTML installed when building packages
> 	@$(MKDIR) -p $(RHOME)/doc/html/search
> 	@$(CAT) $(RHOME)/library/*/CONTENTS >
> $(RHOME)/doc/html/search/index.txt
>
>
> Mark Bravington
> CSIRO Mathematical & Information Sciences
> Marine Laboratory
> Castray Esplanade
> Hobart 7001
> TAS
>
> ph (+61) 3 6232 5118
> fax (+61) 3 6232 5012
> mob (+61) 438 315 623
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>
-- 
Brian D. Ripley,                  ripley at 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 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

> Brian D. Ripley,                  ripley at stats.ox.ac.uk
> This will only happen if you installed R without HTML help 
> and then install a package *with* HTML help *and* do so into 
> a non-standard directory.  That seems an arcane thing to do: 
> why do you want HTML help for some packages and not others?  
> You will end up with a partially functional system.
> 
> The fix is simple: put '-' at the beginning of that line 
> (after the tab) in MakePkg.  I've put a more complete 
> solution in R-devel.
Thanks for the fix. I think the message is more general, though. What I
want to do is install the package just with "vanilla help"-- the stuff
that comes up in the pager. According to "RCMD INSTALL --help" the
only
options are "doc=none/normal/chtml/all"-- there's no
"vanilla" or
equivalent. If I do "doc=none" I don't get vanilla help. So
"doc=normal"
is required, and this tries to do HTML which I don't need. A
vanilla-only option would be ideal, I guess.
> 
> On Mon, 9 Jan 2006 Mark.Bravington at csiro.au wrote:
> 
> > I've just had the error below while trying to install a 
> package from 
> > source under R2.2.1 and Windows XP. I recall encountering this 
> > sporadically in the past. It is a pretty confusing message 
> and took me 
> > quite some time to figure out; I've seen queries about it on the R
<<snipped>>
> > ---------- Making package debug ------------ <<snipped>>
installing
> > indices
> > cat: C:/R/RW2021/library/*/CONTENTS: No such file or directory
<<snipped>>
> > be used? If not, a different error message would be handy.
> 
> The error message is completely informative: why would you 
> want to change it?
I should have said "an additional error message" instead of "a
different...". It's pretty difficult for a user to track down where
&
why it is occurring and whether it matters-- accurate, yes, but not
informative! Anyway, maybe otiose now (the more complete solution hasn't
made it to the R-devel I got this morning, so I can't check yet).
> 
> > I don't understand enough about makefiles to suggest a patch, 
> > unfortunately.
> 
> A good learning project for you?
true no doubt-- but just one out of so many ... :(

Thanks
Mark
> >
> > Mark Bravington
> > CSIRO Mathematical & Information Sciences Marine Laboratory
Castray
> > Esplanade Hobart 7001 TAS
> >
> > ph (+61) 3 6232 5118
> > fax (+61) 3 6232 5012
> > mob (+61) 438 315 623
> >
> > ______________________________________________
> > R-devel at r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-devel
> >
> >
> 
> -- 
> Brian D. Ripley,                  ripley at 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 272866 (PA)
> Oxford OX1 3TG, UK                Fax:  +44 1865 272595
> 
>