(This is a re-post from R-pkg-devel hoping to get more feedback) A package of mine was recently archived from CRAN because one of its examples results in a crash on the M1mac platform: https://www.stats.ox.ac.uk/pub/bdr/M1mac/MeshAgreement.out As I don't have access to a Mac, I'm very grateful for Macbuilder. However, there, the package tests without errors (link likely to stop working): https://mac.r-project.org/macbuilder/results/1670186023-a7f313b721c2b516/ One difference I noticed between the Macbuilder and CRAN systems is that Macbuilder builds for aarch64-apple-darwin20 (64-bit), whereas CRAN builds for aarch64-apple-darwin22.1.0 (64-bit). Moreover, Macbuilder uses R-4.2.1 whereas CRAN uses R-devel r83383. The example leading to the crash uses a function from the imported package cgalMeshes which relies on compiled code. cgalMeshes itself is on CRAN and apparently tests without errors on M1mac. I would like to investigate further how to prevent the crash, but it may depend on the MacOS version, or on the R version. Are there publicly available options to test packages on a system that is very similar to the CRAN M1mac in order to implement the necessary changes to get the package back on CRAN? Many thanks in advance Daniel -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 6923 bytes Desc: not available URL: <https://stat.ethz.ch/pipermail/r-sig-mac/attachments/20221213/29181953/attachment-0001.p7s>
CRAN-like M1mac test environment?
4 messages · Simon Urbanek, Wollschlaeger, Daniel, Uwe Ligges
Daniel, note that you can choose R-devel in MacBuilder (select "development" under R version). If your package works on the MacBuilder then it's good as far as macOS goes since that is identical to the CRAN setup. Cheers, Simon
On Dec 14, 2022, at 12:02 AM, Wollschlaeger, Daniel <wollschlaeger at uni-mainz.de> wrote: (This is a re-post from R-pkg-devel hoping to get more feedback) A package of mine was recently archived from CRAN because one of its examples results in a crash on the M1mac platform: https://www.stats.ox.ac.uk/pub/bdr/M1mac/MeshAgreement.out As I don't have access to a Mac, I'm very grateful for Macbuilder. However, there, the package tests without errors (link likely to stop working): https://mac.r-project.org/macbuilder/results/1670186023-a7f313b721c2b516/ One difference I noticed between the Macbuilder and CRAN systems is that Macbuilder builds for aarch64-apple-darwin20 (64-bit), whereas CRAN builds for aarch64-apple-darwin22.1.0 (64-bit). Moreover, Macbuilder uses R-4.2.1 whereas CRAN uses R-devel r83383. The example leading to the crash uses a function from the imported package cgalMeshes which relies on compiled code. cgalMeshes itself is on CRAN and apparently tests without errors on M1mac. I would like to investigate further how to prevent the crash, but it may depend on the MacOS version, or on the R version. Are there publicly available options to test packages on a system that is very similar to the CRAN M1mac in order to implement the necessary changes to get the package back on CRAN? Many thanks in advance Daniel
_______________________________________________ R-SIG-Mac mailing list R-SIG-Mac at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-mac
Many thanks for your quick response! I apologize for missing the option to select r-devel on Macbuilder. However, the error is not triggered either: https://mac.r-project.org/macbuilder/results/1670933959-5606022c42f1c98d/ Could there be something new about CRAN's aarch64-apple-darwin22.1.0 platform as compared to Macbuilder's aarch64-apple-darwin20 platform? Another thought: In the R-sig-mac archives, I noticed a discussion abou gmp and mpfr. The package I'm using lists gmp and mpfr as SystemRequirements (but does not import the R packages gmp and Rmpfr). Could this be related? Thank you! Daniel
PD Dr. Daniel Wollschl?ger University Medical Center Mainz Institute of Medical Biostatistics, Epidemiology and Informatics (IMBEI) Radiation Epidemiology Tel: +49 6131 17 7029 -----Urspr?ngliche Nachricht----- Von: Simon Urbanek [mailto:simon.urbanek at R-project.org] Gesendet: Dienstag, 13. Dezember 2022 13:17 An: Wollschlaeger, Daniel <wollschlaeger at uni-mainz.de> Cc: r-sig-mac at r-project.org Betreff: Re: [R-SIG-Mac] CRAN-like M1mac test environment? Daniel, note that you can choose R-devel in MacBuilder (select "development" under R version). If your package works on the MacBuilder then it's good as far as macOS goes since that is identical to the CRAN setup. Cheers, Simon > On Dec 14, 2022, at 12:02 AM, Wollschlaeger, Daniel <wollschlaeger at uni-mainz.de> wrote: > > (This is a re-post from R-pkg-devel hoping to get more feedback) > > A package of mine was recently archived from CRAN because one of its > examples results in a crash on the M1mac platform: > https://www.stats.ox.ac.uk/pub/bdr/M1mac/MeshAgreement.out > > As I don't have access to a Mac, I'm very grateful for Macbuilder. > However, there, the package tests without errors (link likely to stop working): > https://mac.r-project.org/macbuilder/results/1670186023-a7f313b721c2b5 > 16/ > > One difference I noticed between the Macbuilder and CRAN systems is > that Macbuilder builds for aarch64-apple-darwin20 (64-bit), whereas > CRAN builds for aarch64-apple-darwin22.1.0 (64-bit). Moreover, > Macbuilder uses R-4.2.1 whereas CRAN uses R-devel r83383. > > The example leading to the crash uses a function from the imported > package cgalMeshes which relies on compiled code. cgalMeshes itself is > on CRAN and apparently tests without errors on M1mac. I would like to > investigate further how to prevent the crash, but it may depend on the > MacOS version, or on the R version. Are there publicly available > options to test packages on a system that is very similar to the CRAN > M1mac in order to implement the necessary changes to get the package back on CRAN? > > Many thanks in advance > > Daniel > > _______________________________________________ > R-SIG-Mac mailing list > R-SIG-Mac at r-project.org > https://stat.ethz.ch/mailman/listinfo/r-sig-mac -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 6923 bytes Desc: not available URL: <https://stat.ethz.ch/pipermail/r-sig-mac/attachments/20221213/34458b38/attachment.p7s>
On 13.12.2022 13:30, Wollschlaeger, Daniel wrote:
Many thanks for your quick response! I apologize for missing the option to select r-devel on Macbuilder. However, the error is not triggered either: https://mac.r-project.org/macbuilder/results/1670933959-5606022c42f1c98d/ Could there be something new about CRAN's aarch64-apple-darwin22.1.0 platform as compared to Macbuilder's aarch64-apple-darwin20 platform?
Yes, could, as the former uses more recent compilers.
Another thought: In the R-sig-mac archives, I noticed a discussion abou gmp and mpfr. The package I'm using lists gmp and mpfr as SystemRequirements (but does not import the R packages gmp and Rmpfr). Could this be related?
Likely the other package links against gmp and mpfr directly if the packages are not used? Best, Uwe Ligges
Thank you! Daniel
_______________________________________________ R-SIG-Mac mailing list R-SIG-Mac at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-mac