[R-pkg-devel] Troubleshooting Winbuilder Run Timeouts

[Andrew Robbins]

Hi again,

I just double checked the test invocations of INMF which are on /much/ 
larger matrices with /far/ more iterations and am still seeing zero 
seconds. That /is/ a bug, but its not a related one. The objective error 
of 44244.4 is what it should be returning for the example matrices.


I just benchmarked devtools::check() (which is actually a build followed 
by a check) on my Ryzen 5700X/intel 660p system with a bunch of stuff 
going on in the background and got just over 11 minutes. I don't 
actually think its timing out. No output differences, of course.


If you could get me more information (perhaps off-list) about the 
-lstdc++ thing I'm curious about reproducing that.

Best.

-Andrew Robbins

On 12/7/2023 5:45 AM, Ivan Krylov wrote:
> On Tue, 5 Dec 2023 17:16:37 -0500
> Andrew Robbins via R-package-devel <r-package-devel using r-project.org>
> wrote:
>
>> Every time, it reports a runtime of zero seconds for INMF res1
> I get exactly 0 seconds for both res1 and res2 on both my Windows
> machines and on my Linux machine. I had discounted that as an integer
> type used somewhere for elapsed time. Is that (exactly 0 seconds and an
> objective value of 44244.4) supposed to indicate a problem?
>
> (On my Linux machine, I had to crudely hack in linking with -lstdc++fs
> because some of the dependencies of RcppPlanc.so didn't pick it up
> automatically.)
>
>> and hangs/is killed on initiation of INMF res2 before the
>> initialization of RcppProgress.
> Can we eliminate it running out of time? On my Windows 10 LTSC machine
> (AMD Ryzen 5 2400G from 2018, SATA SSD, nothing else going on at the
> time) the complete R CMD check takes slightly more than 10.5 minutes.
> An R-release check running on an Intel Xeon E5-2680 from 2016 with
> other checks going on at the same time could plausibly take more than 20
> minutes and get terminated [1], with some of the buffered output
> never reaching the RcppPlanc-Ex.Rout file... but then it happens on a
> much faster R-devel checking machine [2], which runs on an AMD EPYC
> 7443, and we never see any differences in the output. Right?
>
-- 
Andrew Robbins
Systems Analyst, Welch Lab<https://welch-lab.github.io>
University of Michigan
Department of Computational Medicine and Bioinformatics


-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 840 bytes
Desc: OpenPGP digital signature
URL: <https://stat.ethz.ch/pipermail/r-package-devel/attachments/20231207/010ac26f/attachment.sig>