Closed SridharJagannathan closed 4 years ago
Thanks for this @SridharJagannathan. I guess you ran out of steam, when what was supposed to be a simple doc fix turned out to uncover several other problems. I think that most of these are related to changes in behaviour of R 4.0 (stringsAsFactors) / (https://github.com/r-lib/devtools/issues/2216) in which tests in \donttest
blocks are now run by default!
Thanks @jefferis I tried your suggestion of setting the env var but it still fails (infact in one of my previous commits i tried changing \donttest to \dontrun, which should do the same thing as the env var but it still failed), so the issue is not only due to the \donttest block but due to warnings being converted to errors by the texlive packages, the error message was a bit cryptic at that time so i decided to come back to it later (https://travis-ci.org/github/natverse/rcatmaid/jobs/696263033) you can see example here: (https://devops.stackexchange.com/questions/8014/travis-tests-for-r-package-fail-with-little-explanation). I have now changed that behaviour by using warnings_are_errors: false
which lets the travis build pass, but I guess the larger problem is investigating the texlive issues.
Hi @SridharJagannathan, I suspect that the remaining bug is not to do with texlive but actually in a different place, here:
https://travis-ci.org/github/natverse/rcatmaid/jobs/696269530#L968-L969
I haven't figured out the fix but it suggest some kind of code/documentation mismatch for this AV4b1 object.
@jefferis The texlive issue seems to be present for other users too: https://travis-ci.community/t/build-failure-due-to-latex-font-warning-font-shape-t1-zi4-m-it-undefined-fixed-with/3238 which hasn't been yet resolved.
Possibly but the issue with the AV4b1 is definitely a warning in R check (i.e. an error by default). We can sidestep the tex issue by just not building the PDF manual on travis. There is a travis yml setting for that. Like so, I think.
r_build_args: --no-manual
r_check_args: --no-manual
I think adding
#' @usage data(AV4b1)
in catmaid.R
should fix that other warning.
OK. I think you've cracked it @SridharJagannathan! Thanks so much, G.
Thanks @jefferis fixed the issue wrt AV4b1
and reverted to warnings_are_errors: true
, the build passes now.
Added a couple more example fixes. Now donttest tests work fine on my local machine using the l1em.virtualflybrain.org as target.
Updated for #158