[R-pkg-devel] Mega2R CRAN submission: false positive

[Kevin Ushey]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120837 suggests it's a
compiler bug; hopefully the issue will go away once the version of gcc
used on these CRAN check machines is updated.

On Thu, Jul 24, 2025 at 8:10 AM Weeks, Daniel E <weeks using pitt.edu> wrote:
>
> Last week I saw that someone requested help because expected e-mails from CRAN regarding an R package submission were not received.  I am afraid I may be in the same boat regarding trying to get our R package Mega2R fixed so it can be in CRAN again, as I received no response regarding my pointing out that the automated rejection of Mega2R is a false positive due to a compiler bug that is beyond my capability to fix.
>
> I submitted Mega2R on July 14, 2025, and received an automated message that
>
> package Mega2R_1.2.0.tar.gz does not pass the incoming checks automatically.
>
> I responded as below that it was a false postive but did not receive any response after that.
>
> --------------
>
> I believe this rejection is a false positive due to a subtle compiler error.
>
> In June, our Mega2R package was rejected due to a Warning that was generated during the CRAN checks. The Warning (which we are stilll seeing) was:
>
> ```
> Flavor: r-devel-linux-x86_64-debian-special-gcc-san
> Check: Post-processing issues found for gcc-san, Result: WARNING
>    File: build_vignettes.log
>    vendor/sqlite3/sqlite3.c:80239:14: runtime error: load of address
> 0x7faa31b1fa40 with insufficient space for an object of type 'struct
> MemPage *'
> ```
>
> As I explain below, I think this Warning is not due to any mistake in Mega2R code itself, but rather due to a compiler bug that is rarely encountered only under specific settings.  As such, I hope you will approve releasing Mega2R to CRAN based on the following due diligence investigations and findings:
>
> Note that the Warning is generated by sqlite3 code so I opened a thread about it on the SQLite forum:
>
> https://sqlite.org/forum/forumpost/e1015d1714
>
> The conclusion of one investigation into this concludes that
>
> "Because of the requirements to reproduce the issue, it looks like a compiler issue not a bug in SQLite"
>
> Further checks of Mega2R
>
> The Warning seen on the gcc-san test machine is not
> reproducible on the other similar test machines. The warning is not reproducible on the other machines.
>
> Further investigations via rhub
>
> https://github.com/r-hub2/diplomatic-mole-Mega2R/actions
>
> None of the checks on these machines reproduced the Warning seen on the gcc-san test machine:
>
> clang-asan: All CRAN checks passed.
>
> clang-ubsan: All CRAN checks passed.
>
> valgrind: All CRAN checks passed.
>
> m1-san (R-devel): The Warning seen on the CRAN gcc-san test machine was not observed.  However, there was an error involving `seqlevels<-` (which is also observed using R Under development (unstable) (2025-07-11 r88405)).  This is because the development version of `GenomeInfoDb` no longer contains the `seqlevels<-` function, which in the next release
> will be moved into the Seqinfo package.  At that time, this Roxygen command
>
> ```
> #' @importMethodsFrom GenomeInfoDb 'seqlevels<-'
> ```
>
> will need to be updated.  For now, in the currently released versions of R such as R version 4.5.1 (2025-06-13), this is correct code.
> ------------------------
>
>
>
> Thanks,
>
> -- Dan --
>
> Daniel E. Weeks, Ph.D.
> Professor of Human Genetics
> Professor of Biostatistics and Health Data Science
> Department of Human Genetics
> School of Public Health
> University of Pittsburgh
> Public Health 3119
> 130 DeSoto Street
> Pittsburgh, PA 15261
> USA
>
> (412) 624-5388
> FAX: (412) 624-3020
>
>
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-package-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-package-devel