[R-pkg-team] Bug#1037439: Opinions about removal of 32bit architecture for r-cran-rstan (Was: Bug#1037439: r-cran-rstan/armhf FTBFS with r-cran-bh 1.74)

[Andreas Tille]

Hi Adrian and Dirk,

thanks a lot to Adrian for taking the time to dig into R source code for
a fix.

Am Mon, Jun 19, 2023 at 12:04:02PM -0500 schrieb Dirk Eddelbuettel:
> | > Not sure that it is fair to point at BH / Boost though.  Anyway.
> | > 
> | > CRAN no longer checks / compiles 32 bit so upstream may not care, but they
> | > are a good team (if busy).  You could ping Ben, he is at
> | >    Benjamin K Goodrich <benjamin.goodrich at columbia.edu>
> | >...
> | 
> | The patch below to r-base fixes the build of r-cran-rstan/i386 for me.

I really welcome your effort in this bug.  Thanks again.
 
> Shucks:  You missed a new upstream by a few days.  I'll consider a -2 but
> note that ...
>  
> | This will reduce debug info in the R ecosystem on 32bit to what is 
> | required for backtraces, but I assume realistically R on 32bit is
> | anyway only sparsely used these days.

>From all user responses I've got so far none of them is using 32bit (for
years).

> ... 32 bit is no longer a release architecture for R upstream so this is
> borderline either way.

>From my point of view it is simply burning CPU cycles ouf our
autobuilders network if we stick to the probably only theoretical option
a user might use this package on some 32bit architecture.  So I tend to
simply drop 32bit support for r-cran-rstan (and packages that depend
from it which will end up in a bunch of removal bugs for ftpmaster but
if I'm honest on my "save the environment and stop 32bit support for
this" I need to do this).

If someone thinks we should keep 32bit support and apply the patch
suggested to r-base (which Dirk seems to be willing to) please speak
up soonish (in the next two weeks).

Kind regards
    Andreas.


-- 
http://fam-tille.de