Hi Yavor, > Étienne Mollier wrote: > > Thank you for tackling all those issues on the ball package, I > > have reviewed your changes and I consider the package to be fit > > for upload as-is. I will proceed to a sponsored upload soon. > > Thanks but another upload is needed because of my autopkgtest snafu > (#1120691) that I should have really predicted.
No worries, part of my sponsoring work should probably have covered less usual processor architectures, but that would have multiplied the amount of tests I would have run more or less by hand. It's good that our CI caught the issue in the end; too bad it required human intervention from Paul Gevers due to the memory saturation condition; hopefully CI runners will guard against excessive memory usage in a near future. I think we can thank him for his work on the CI infrastructure. Back to your changes to ball, I reviewed your work and verified that your reduced variant of autopkgtest had reasonable memory consumption on ppc64el and s390x. I had cruise memory usage of about 10GiB of RAM, but I believe that: first most of it seemed imputable to my virtualisation layer, second there was a peak at about 15GiB which seems to be constrained to a short time period at the beginning of the processing, but it seems not sufficient anymore to strain CI runners. I have also otherwise a lot of activity on my machine, so something else might have been interferring, and investigating with top didn't raise outstanding memory usage from the ball binaries. It is great that the package is cross buildable, because this really sped up my review in this particular situation. I reviewed also your other changes and have no objection to them. ball 1.5.0+git20220524.d85d2dd-2 is therefore uploaded. Have a nice day, :) -- .''`. Étienne Mollier <[email protected]> : :' : pgp: 8f91 b227 c7d6 f2b1 948c 8236 793c f67e 8f0d 11da `. `' sent from /dev/pts/2, please excuse my verbosity `- on air: Renaissance - The Harbour
signature.asc
Description: PGP signature
