[R-pkg-team] Bug#1103118: r-cran-timechange: should depend on R>=4.5

[Chris Hofstaedtler]

* Chris Hofstaedtler <zeha at debian.org> [250429 17:24]:
>* Rebecca N. Palmer <rebecca_palmer at zoho.com> [250429 17:21]:
>>This isn't a 32-bit problem as such - it's another package that 
>>picked up the R_getVar symbol (new in R 4.5) when it was binNMUd on 
>>some architectures.
>>
>>(The other known one is r-cran-svglite #1103369.  Like that one, the 
>>best we can do after the fact is to add an explicit R>=4.5 
>>dependency, which I am now doing.)
>
>BTW, why is this dependency not automatically generated? Shouldn't 
>this be fully automatic with dh_shlibdeps?

Weird enough: r-cran-timechange does not directly depend on 
r-base-core. It depends on r-api-4.0, which is a virtual package 
provided by r-base-core.

I don't really know what exactly is guaranteed or encompassed by 
"r-api-4.0". Maybe r-base-core should really provide r-api-4.5 
instead? How is this all supposed to work?

dh_shlibdeps is called during the r-cran-timechange build, but it 
seems to stay quiet on timechange.so? :-|

Chris