[R-SIG-Mac] Failed tests on M1mac

[Prof Brian Ripley]

On 22/11/2025 02:37, Marc Schwartz via R-SIG-Mac wrote:
> On Nov 21, 2025, at 9:21 PM, Duncan Murdoch <murdoch.duncan using gmail.com> wrote:
>>
>> On 2025-11-21 6:53 p.m., Jordan Martinetti wrote:
>>> Hello I am a new CRAN author working on Windows.  I tested my package locally and it worked great and passed CRAN testing when I initially uploaded.  Yesterday I got an email saying it failed on M1mac.  I don't have a mac machine to do testing on so i'm unsure how I can test my package before uploading to CRAN and hoping it passes.
>>> While I don't expect a significant portion of users of my package to work on mac, I would rather not skip mac testing and try to work on as many platforms as possible.  How can I test my package on mac, without owning a mac, so that I know it will work before uploading to CRAN?

You do not have the option to exclude Mac users under the CRAN policy, 
and this is in any case not a Mac issue.

>> If you use Github actions for testing, you can choose a Mac machine to run on.  I don't know what versions of macOS they have available.
>>
>> Here are some lines from the YAML file for one of my packages that is tested on Windows, Ubuntu and macOS:
>>
>>   R-CMD-check:
>>     runs-on: ${{ matrix.config.os }}
>>
>>     name: ${{ matrix.config.os }} (${{ matrix.config.r }})
>>
>>     strategy:
>>       fail-fast: false
>>       matrix:
>>         config:
>>           - {os: windows-latest, r: 'release'}
>>           - {os: windows-latest, r: 'devel'}
>>           - {os: ubuntu-latest,   r: 'devel', http-user-agent: 'release'}
>>           - {os: macos-latest, r: 'devel'}
>>
>> These were originally based on the code from the `usethis` package; I don't know how far it has diverged by now.
>>
>> If you're not using Github, Gitlab or something similar, you should.
>>
>> Duncan
> 
> 
> Hi,
> 
> In addition to Duncan's suggestion, you can use the macOS Builder service here:
> 
>    https://mac.r-project.org/macbuilder/submit.html
> 
> It generally parallels the winbuilder site (https://win-builder.r-project.org/) for the testing of CRAN packages on Apple silicon.
> 
> It is listed in the Re-submission section in the CRAN Policy:
> 
>    https://cran.r-project.org/web/packages/policies.html#Re_002dsubmission

That would be a good answer if this were a macOS issue, but the OP 
jumped to that conclusion and it was not reported to him as such.

This appears to be about package sodaR and the additional issue showing 
for thqt is about a date-time (and it passes on other macOS platforms). 
That suggests that it depends on time zone or locale -- the OP uses a 
test-obfuscation suite but digging into the package sources shows that 
this is attempting to work with

"09/14/2012 10:38:01 PM"

The help for posixify has no examples and no description of its expected 
inputs.  It claims the result is 'a POSIX date', which is false (it 
appears to be a date-time of class "POSIXct").

That is both a country-specific format and locale-dependent, as pointed 
out in the help for strptime.  In this case locale-specific conversions 
depend not just on the locale name but also the OS version -- macOS has 
changed its locale definitions since version 11 used for the CRAN macOS 
checks (and AFAIK for mac-builder).

If you really need to convert that benighted format more portably, you 
could try setting a C locale by e.g.

Sys.setlocale("LC_TIME", "C")

The POSIX standard mandates AM/PM strings in the C locale but OSes do 
not necessarily pay close attention to such standards.


-- 
Brian D. Ripley,                  ripleybd using icloud.com
Emeritus Professor of Applied Statistics, University of Oxford