El 14/3/25 a las 6:58, Michael Tokarev escribió:
Yes, this is a known issue, I discovered it right after uploading 3.13.0-1, notified upstream and we come up with the fix you added. Since this prob only happens when doing non-parallel build (as you noted too), I didn't bother uploading -2 version..
Hi. To be precise, I didn't say that it only happened on single-cpu systems. What I said is that on single-cpu systems, the build seemed to always fail. However, being a Makefile bug, we can't be sure in which systems it will build and which ones it will not. It could be, for example, that it fails to build if the number of CPUs is odd, who knows? For me who try to ensure that the end user will be able to build packages in their systems, it is unfortunate that (severities aside) people seem to believe that building ok in the buildds is good enough for a stable release. The end user who want to build the package from source will not do so in the buildds, they will do that in their own computers, but also using virtual machines in the cloud, where single-cpu systems are more cost-efficient.
but you filed this bug. I'm sorry for the work you had to do here, for an issue which is well-known to the maintainer..
Well, there is also this side comment which I did, for which I didn't get any answer:
Note 2: After fixing the build problem please consider fixing the failing test that will happen afterwards as well, as it also fails 100% of the time here.
Admittedly, I should have filed a different bug for that. Please tell me: Can you (or anybody in the team) build the package with GRUB_CMDLINE_LINUX="nr_cpus=1", or you get failing tests as I do? I'd like to know that you can reproduce the bug before I report it. (In either case, I would offer a VM to reproduce, as I always do). Thanks.
