[R-pkg-team] Bug#1064135: bigmemory: fails to migrate to testing for too long: autopkgtest fails on arm*, ppc64el and s390x

[Charles Plessy]

On Thu, 27 Jun 2024 11:21:44 +0200 Andreas Tille <andreas at an3as.eu> wrote:
> 
> [1] https://github.com/kaneplusplus/bigmemory/issues/115

Hi all,

the name of the failing test is 'Making sure the 'char' matrix behaves 
like signed byte'.  The patch on Ubuntu that fixes the issue replaces 
the expected value by another one, but I am not sure if this just hides 
the dust under the carpet…

For the record I just have posted that on the GitHub issue:

_bigmemory_ passes the tests on CRAN's `arm64` platforms, which all run 
on MacOS. On Linux it fails on `arm64` and a lot of other architectures. 
  In my understanding, `char` is signed on `macos-arm64` and linux 
`amd64` (`x86_64`) and unsigned on linux `arm64`. Would it solve the 
problem if _bigmemory_ would pass `-fsigned-char` to the compilers?

Cheers,

Charles

-- 
Charles Plessy - - ~ ~ ~ ~ ~ ~~~~ ~ ~ ~ ~ ~ - - charles.plessy at oist.jp
Okinawa  Institute  of  Science  and  Technology  Graduate  University
Senior staff scientist ~~ Luscombe Unit ~~ https://groups.oist.jp/grsu
Toots from work - ~~ ~~~~ ~~ - https://fediscience.org/@charles_plessy