On Nov 29, 2021, at 5:51 PM, Herv? Pag?s <hpages.on.github at gmail.com> wrote:
On 29/11/2021 09:25, Zhou, Wanding wrote:
Thanks for the input Herve. I saw your post but can you help me confirm that mine is on the 32-bit windows since it says mingw32 (64-bit)?
Windows Server 2012 R2 Standard / x64
* using platform: x86_64-w64-mingw32 (64-bit)
This shows the Windows arch that R uses by default on Windows. But we use --force-multiarch when running 'R CMD check' on our Windows builders to force 'R CMD check' to run all the examples, unit tests, and vignettes in the two supported Windows archs: 32-bit and 64-bit.
The error was happening during the following step:
** running examples for arch 'i386'
and not during the following step:
** running examples for arch 'x64'
So yes, this is a Windows 32-bit only error.
I still don't think I am using that much but I will remove some examples and see that solves the problem
Here is one way to see how much memory 'R CMD check' consumes when running the examples:
1. Run 'R CMD check' on the package. This collates all the examples in sesame.Rcheck/sesame-Ex.R.
2. Comment out the last line (quit('no')) in sesame.Rcheck/sesame-Ex.R.
3. From R do: source("sesame.Rcheck/sesame-Ex.R", echo=TRUE). At the same time, monitor memory usage in another terminal e.g. by running 'top' on Linux/macOS.
4.Once all the examples have finished, do gc().
This is what I get on my laptop when doing the above with sesame 1.13.3:
used (Mb) gc trigger (Mb) max used (Mb)
Ncells 15062911 804.5 24349446 1300.5 24349446 1300.5
Vcells 278380669 2123.9 473158124 3610.0 473157073 3610.0
That's a memory consumption of 4.9G which is confirmed by the 'top' command I was running in another terminal.
It would be awesome if 'R CMD check' could do the above for us i.e. report the amount of memory used by the examples. Any volunteer to request this feature on the R-devel mailing list?
Hope this helps,
H.
Hi,
Recently we are experience some memory issue on both development and release (mostly Windows) that causes the sesame package build to fail. We are trying to address it but it seems to run fine locally on our side.
This is the error on the development branch
Setting options('download.file.method.GEOquery'='auto')
Setting options('GEOquery.inmemory.gpl'=FALSE)
Error: processing vignette 'sesame.Rmd' failed with diagnostics:
C stack usage 7977908 is too close to the limit
And this is the error on the release branch
data = data.frame(estimate=c(runif(10, 0, 10)))
plotLollipop(data)