Closed rdiaz02 closed 6 years ago
Thanks. Can we safely change --no-multiarch
to --merge-multiarch
? It would be great if you could try this in a fork.
Sorry, I am not sure to understand. I cannot try to use --merge-multiarch
because appveyor seems to have no-multiarch
hardcoded.
If the question is whether the build system in appveyor can use --merge-multiarch
well... I am not sure I follow the intricacies of the thread in http://r.789695.n4.nabble.com/R-CMD-check-force-multiarch-does-not-install-all-the-archs-for-testing-td3629129.html But BioConductor seems to use
--merge-multiarch
for the R CMD INSTALL --build
step--force-multiarch
for the R CMD check
step(This is an example http://bioconductor.org/checkResults/devel/bioc-LATEST/OncoSimulR/tokay1-checksrc.html)
So my understanding (again, take with many grains of salt) is that allowing something like the following should work:
R_INSTALL_ARGS: --merge-multiarch
R_CHECK_ARGS: --force-multiarch
and the problem seems to arise because I cannot use anything like the first, as INSTALL_opts="--no-multiarch"
is hardcoded.
Thanks. I was suggesting that you fork r-appveyor and try the changes you propose there on your project. If this works I'm happy to integrate them in this repository.
As for the hard-coded value, we might just need to provide one or two new environment variables which allow overriding the --no-multiarch
setting.
Thanks! It is working fine now, though. I forked the repo and I'll try what you are suggesting when it fails again. The problem seems to be building some packages from source: now that the binaries are available, things work again.
We are using appveyor to generate windows packages that we put straight into drat, so it would be nice for biarch builds in that use case (eg https://github.com/DeclareDesign/randomizr/issues/46).
I forked this repo, edited out the --no-multiarch
here (https://github.com/nfultz/r-appveyor/commit/0ed58534af1a5c61387891c90eac409a69e77e6b) and got a clean build: https://ci.appveyor.com/project/DeclareDesign/randomizr/build/job/fb09vel0ujow41gm
It sucessfully made a biarch package that got placed in to drat: https://github.com/DeclareDesign/declaredesign.github.io/blob/master/bin/windows/contrib/3.4/PACKAGES#L33
@nfultz: Thanks, would you like to submit a PR?
/remind me on Tuesday
@krlmlr set a reminder for Mar 6th 2018
:wave: @krlmlr,
I am getting errors in a package that I think can be traced to a problem I have not seen in the past with RBGL.
When my
appeyor.yml
file hastravis-tool.sh install_bioc RBGL
it seems to run successfully as I eventually see (https://ci.appveyor.com/project/rdiaz02/OncoSimul/build/job/0wwqwkbq5ldw0qai) (this is for a x64 build; similar but referring to i386 in the i386 build)But I see the message
And then, further down below, when another package (
nem
in this case) that depends on RBGL has to be installed I seeIf I do not directly try to install RBGL and leave it for the dependencies to be resolved, the same thing happens (https://ci.appveyor.com/project/rdiaz02/oncosimul/build/1.0.372/job/5837y47575gax7mi) .
It seems I would need to use the proper magical combination of
--merge-multiarch
and/or--force-multiarch
(http://r.789695.n4.nabble.com/R-CMD-check-force-multiarch-does-not-install-all-the-archs-for-testing-td3629129.html) but I have not been able to get it to work, as appveyor seems to have theINSTALL_opts="--no-multiarch"
harcoded.Any possible workarounds?