Hi, I'm just trying to follow along with the instructions in the readme.

I tried install.packages("rhub2") but it said my version of R is not compatible.

So I installed with devtools::install_github("r-hub/rhub2") and then followed the steps to make a yml file and commit it to github.

Next I tried the rhub_doctor() command and got an error:

> rhub_doctor()
✔ Found R package at /path/work/github.com/slowkow/ggrepel.
✔ Found git repository at /path/work/github.com/slowkow/ggrepel.
✔ Found GitHub PAT.

Error in if (pcs$host == "github.com") { :
  missing value where TRUE/FALSE needed

> traceback()
5: parse_gh_url(gh_url)
4: doctor_async_gql(gh_url, token = pat)
3: when_all(gql = doctor_async_gql(gh_url, token = pat), wfl = doctor_async_rest(gh_url,
       token = pat))
2: synchronise(when_all(gql = doctor_async_gql(gh_url, token = pat),
       wfl = doctor_async_rest(gh_url, token = pat)))
1: rhub_doctor()

I need to build a Mac OSX binary for a user of the deldir package. But there is no longer any trace of "macos" when I execute platforms().

Earlier today (NZ time) I thought I saw mention of this problem on github, indicating that macos builds were suspended because they were not working, But I can no longer find anything about this. There is nothing that I can see in "Issues".

Did I imagine seeing this mention?

Will macos builds come back to life?

On demand. Maybe needs a config option? The TeXlive packages also need a config option.

These config options (and others as well) could default to some Config/GHA entries in DESCRIPTION.

Need to improve the cache key for R builds/packages that

• link to libc++ on Linux
• don't have an R shared library

Hello, here Christian Steglich from the RSiena package for longitudinal network analysis. I was happy to come across the rhub2 package and would like to thank you for providing this service in the first place!
For our package https://github.com/snlab-nl/rsiena we recently got a complaint of memory leaks diagnosed by valgrind from the CRAN people. As Windows user, I cannot work with valgrind to replicate this locally, so tried it via rhub2. It appears that the valgrind checks do not detect anything wrong, unlike the memchecks run at CRAN. I wonder why. Does this mean that rhub2 doesn't check for memory leaks? If not, could it be made to do that?
Additionally, we do not get the protocol files generated when checking via rhub2. Because these contain additonal detailed information when running the checks via rhub2. Is it possible to get them? Edit: I see that this latter aspect may be what Issue #8 refers to.

Always? Should be fast enough.

For flexibility.

Hi, this looks like a fantastic package that will be very useful in package development!

I've tried to set it up for one of my packages, which lists a Suggested package ("TreeDistData") using the Additional_repositories: description field. This causes the default workflow runs to fail:

  Error: 
  ! error in pak subprocess
  Caused by error: 
  ! Could not solve package dependencies:
  * deps::.: Can't install dependency TreeDistData (> 0.1.0)
  * TreeDistData: Can't find package called TreeDistData.
  * any::sessioninfo: dependency conflict
  * any::rcmdcheck: dependency conflict

To some extent this represents an issue with pak, but I've found it possible to work around this bug in my existing GHA workflows by installing "TreeDistData" from GitHub.
I'm wondering whether I can configure rhub2 to do the same thing; it's not clear that I can achieve this by editing rhub.yaml. Would it be possible to support an option under the with: field to, for example, the r-hub/rhub2/actions/rhub-check action?

Etc. and also fail the check on failures.