[R-pkg-devel] Segfault error in CRAN submission (Linux)

[Dirk Eddelbuettel]

On 29 September 2025 at 19:14, quirin stier wrote:
| I submitted a locally functioning package to CRAN (see 
| https://github.com/Quirinms/GeneralizedUmatrixGPU), however in the case 
| of the Linux test, I get following error:
| 
|   *** caught segfault ***
| address 0x10, cause 'memory not mapped'

I would recommend investing some time in a (possibly Docker-based) setup that
closely matches the system where this failed so see if you can replicate it
so that you can try to find ways to avoid it.

For what it is worth, I did a quick-ish test in a r2u-based container (based
on Ubuntu 24.04), installed your dependencies (as binaries apart from
GeneralizedUmatrix) as well as this package. It passes its regression tests
and did not segfault.

That does not prove correctness. The next step is then often to try tools
like valgrind or ASAN/SAN, also in containers, to get other hints.  I have
been this for a number years (decades actually) and often thought "well CRAN
must be wrong here". They never were. The tests are there for a reason, even
if it pains us.

(I would also think about changing the language on your webpage. Referring to
two packages as CRAN packages when they are not does not help casual users.)

Cheers, Dirk

-- 
dirk.eddelbuettel.com | @eddelbuettel | edd using debian.org