[R-pkg-devel] Weird error on CRAN linux check

[Iñaki Úcar]

El jue., 5 jul. 2018 21:51, David Hugh-Jones <davidhughjones using gmail.com>
escribió:

> That will indeed fail everywhere.  The puzzle is why it fails (only
> sometimes) when the methods are all exported. The GitHub equivalent is tag
> v4.0.1-rc1.
>

Yeap, but the point is that exporting the methods with 'export' should be
superfluous if S3method is used.


> On Thu, 5 Jul 2018 at 20:43, Iñaki Úcar <i.ucar86 using gmail.com> wrote:
>
>>
>>
>> El jue., 5 jul. 2018 21:35, David Hugh-Jones <davidhughjones using gmail.com>
>> escribió:
>>
>>> Installed from CRAN or github? CRAN should be OK - I hope!
>>>
>>
>> From GitHub before the patch.
>>
>>
>>> On Thu, 5 Jul 2018 at 20:33, Iñaki Úcar <i.ucar86 using gmail.com> wrote:
>>>
>>>> I installed huxtable in two environments, my own Fedora installation
>>>> with R 3.5.0 and all my packages and in a fresh Ubuntu system with R 3.4.4
>>>> and an empty library. In both cases, huxtable is unusable: every example I
>>>> try fails because it doesn't find the methods.
>>>>
>>>> So it has nothing to do with R checks or CRAN scripts, and it seems
>>>> improbable to me that the error comes from a corrupted dependency.
>>>>
>>>> Iñaki
>>>>
>>>>
>>>> El jue., 5 jul. 2018 20:06, Duncan Murdoch <murdoch.duncan using gmail.com>
>>>> escribió:
>>>>
>>>>> On 05/07/2018 9:11 AM, David Hugh-Jones wrote:
>>>>> >
>>>>> > Agreed. I fixed the roxygen2 and it works fine. But yet, the
>>>>> original
>>>>> > v4.0.1 on CRAN has a namespace file which contains
>>>>> >
>>>>> > S3method(bold,huxtable)
>>>>> > export(bold)
>>>>> > export(bold.huxtable)
>>>>> >
>>>>> > and
>>>>> >
>>>>> > S3method("align<-",huxtable)
>>>>> > export("align<-")
>>>>> > export("align<-.huxtable")
>>>>> >
>>>>> > yet still fails on linux-patched and linux-release, with "no
>>>>> applicable
>>>>> > method" errors for align<- and bold. Unfortunately, I don't know how
>>>>> to
>>>>> > reproduce the error on any other platform....
>>>>>
>>>>> I just got R installed on an Ubuntu VM, and ran "R CMD check
>>>>> huxtable_4.0.1.tar.gz" both with and without "--as-cran", without
>>>>> seeing
>>>>> the error you quoted.  (I did see other problems, related to not
>>>>> having
>>>>> things like pandoc installed; nothing that looked like a problem with
>>>>> the package rather than a problem with my R installation.)
>>>>>
>>>>> That looks like a bug, but without having a system that can reproduce
>>>>> it, it's hard to narrow down where:
>>>>>
>>>>>   - In R's checks?  Seems unlikely, given it is so system specific.
>>>>>   - In CRAN's scripts?  Really unlikely, since all the tests are in R.
>>>>>   - In huxtable or some package used by huxtable?  Seems possible:
>>>>> maybe memory got corrupted.  Perhaps running under some memory checker
>>>>> would be more informative.
>>>>>
>>>>> Perhaps the CRAN team could be helpful here.
>>>>>
>>>>> >
>>>>> > Anyway, meanwhile, my problem is fixed and I have learned something
>>>>> > about function environments.
>>>>>
>>>>> Given that the error is unrelated to the solution, it really looks
>>>>> like
>>>>> memory corruption somewhere or other.
>>>>>
>>>>> Duncan Murdoch
>>>>>
>>>> --
>>> Sent from Gmail Mobile
>>>
>> --
> Sent from Gmail Mobile
>

	[[alternative HTML version deleted]]