[Rd] [External] Re: R 4.0.0 build error with sysdata.rda on ppc64el architecture

[Tomas Kalibera]

On 4/30/20 4:37 PM, Dirk Eddelbuettel wrote:
> On 30 April 2020 at 09:11, luke-tierney using uiowa.edu wrote:
> | On Thu, 30 Apr 2020, Dirk Eddelbuettel wrote:
> | Maybe I missed something. How is the 'compiler' package involved?
>
> See the other email thread; you replied (~ 26 hours ago) to my message adding
> that "sysdata.rda in 'tools' hanging was wrong". It gets to the next stage,
> which is 'compiler'.
>
> We know by now (thanks to Bryan and Tomas) that src/main/machar.c is the real
> culprit due to the long double behaviour on ppc64el.  I presume 'compiler'
> calls into platform.c and hence machar.c for calibration or setup.

Yes, R just hangs during startup, always, it is not related to the 
compiler nor sysdata.rda.
The problem is that machine characteristics detection does not work with 
long double on ppc64el, it falls into an infinite loop.

Tomas

>
> On the Debian side, I just shipped an updated source package which will skip
> long double support on this platform.  That should allow builds to complete,
> hopefully, so that we can look into get the build from 'experimental' into
> 'unstable' and the distro proper.  In the meantime, there are alternative
> venues for getting R 4.0.0 binaries for Debian and Ubuntu as previously
> mentioned.
>
> Dirk
>