With `BuildVignettes: no` in my DESCRIPTION and `^vignettes` in my
.Rbuildignore, I can squeak through a normal build (no `r_build_args` or
`r_check_args`) on travis with just this note:
* checking for old-style vignette sources ... NOTE
Vignette sources only in ‘inst/doc’:
‘rslurm.Rmd’
A ‘vignettes’ directory is required as from R 3.1.0
and these will not be indexed nor checked
I’ll take a NOTE over an ERROR, so I’ll try to submit like this with an
explanation. It seems to me like `BuildVignettes: no` should also disable
re-building them under R CMD check, but I think R CMD check --as-cran clobbers
it.
-Original Message-
From: Dirk Eddelbuettel on behalf of Dirk
Eddelbuettel
Date: Thursday, April 6, 2017 at 9:23 PM
To: Ian Carroll
Cc: "r-package-devel@r-project.org"
Subject: Re: [R-pkg-devel] maintainer built vignette
On 6 April 2017 at 20:17, Ian Carroll wrote:
| How do I package an rmarkdown vignette that cannot be built on
win-builder? The vignette in question demonstrates interaction with a SLURM
cluster, which is not available to win-builder or travis. In my `.travis.yml` I
can include `r_build_args: "--no-build-vignettes"` and `r_check_args:
"--no-vignettes"` to skip the build and re-build steps respectively. I don’t
know how to do that with win-builder, so I cannot pass the incoming pretest on
package submission. The extra steps make me feel like I’m doing something wrong
regardless.
|
| So, what’s the appropriate way to include vignettes that should not be
hammered with attempts to build at every turn? Some old advice seemed to
suggest going straight to inst/docs, but newer rules are that inclusion in
vignettes/ is mandatory.
I think if you adjust .Rbuildignore to NOT include the source but just the
pdf then you will have documentation present -- without the risk of 'cannot
be computed here' on some platforms.
Dirk
--
http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
__
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
