Re: Removal of S.h (was: Re: R 4.2.0 suggests a rebuild of graphics packages)
Hi Johannes,
Am Mon, Apr 25, 2022 at 11:22:11AM +0200 schrieb Johannes Ranke:
> Am Montag, 25. April 2022, 10:35:57 CEST schrieb Andreas Tille:
> > I agree that we have two different issues. However, since we need an
> > r-api-* bump anyway (IMHO)
>
> I was not aware that you are suggesting an API bump for R 4.2.0, for me this
> was not clear from your mail from your 13 April to debian-r.
I admit I was on semi-vac and thus may be a bit sloppy with my wording.
> Regarding the removal of S.h, the R 4.2.0 NEWS file states:
>
> * The header S.h which has been unsupported since Jan 2016 has been
> removed. Use R.h instead.
>
> This suggests that R core sees this as a bug in (bioconductor) packages still
> using S.h, and not as an API change.
I confirm that I see the bug in the bioconductor packages. However, it is
a given fact that the version bump of R to 4.2.0 makes (at least) 4 packages
FTBFS and we have > 150 rdepends of these packages.
> It seems to me, the situation being as it is, the easiest way to fix
> bioconductor in unstable and testing is to upload the new versions scheduled
> for April 27 before r-base migrates to testing. Not exactly clean, but bumping
> the API now would require managing a transition and rebuilding hundreds of
> packages that do not have an issue, right?
Yes. However, codesearch uncovers at least one non-BioC package:
r-cran-locfit_1.5-9.5-1/src/local.h
#ifdef SVERSION
#include <S.h>
typedef long int Sint;
typedef s_object * SEXP;
This is dealt with by `#ifdef SVERSION` which seems to be not set on Debian
but I'm not sure whether this has catched all occurences. I'm not sure what
side effects we might meet if we do not rebuild.
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: