Crashes on ampere1 have been an issue for months. And not just 142arm64-quarterly. 134arm64 and 134releng-armv7 were past examples.
Most evidence related to crashes is not publicly available, so far as I know. ampere1 is now trying a 6th time for 142arm64-quarterly, but this time only having 8948 to build of 10782 queued. So: significantly less queued so it has a better chance of finishing the specific build. Note: the 6 attempts updated the ports tree between attempts. Unlike some past problems, I've not noticed any obvious patterns across what was listed at the times of the crashes. That would suggest not being a problem with any specific port-package or type of port-package. If so, the issue may be too rare to readily investigate and difficult to produce in a development context. Yet is preventing timely quarterly port-package distribution for aarch64 (tier 1), not just armv7. So what does FreeBSD do in such cases when a tier 1 platform (aarch64) is involved? Note: I'm not aware of such crash problems for: ) ampere3 (135/142 arm64/armv7 latest builder) ) ampere2 ( main arm64/armv7 latest builder) It is not obvious to me why quarterly would be a significant difference and that might suggest ampere1 itself is contributing to the failures, not just the software. === Mark Millard marklmi at yahoo.com
