Hi all, so let me shortly explain how that GCC testing with ia64 hardware works:
For every Linux mainline RC and release (usually published between Sunday and Monday) I built three versions of the kernel based on a single kernel configuration for real hardware plus at least one extra for the HP Sim platform running in Ski based on the default kernel configuration for HP Sim. Three versions because: 1. one version built with the last known-good gcc-14 snapshot (since 2024-03-04 always w/LRA enabled!) 2. one version built with the current gcc-14 snapshot (w/o LRA enabled) 3. one version built with the current gcc-14 snapshot (w/LRA enabled, kernel version ends with "-lra"!) These are then (boot to login) tested on the following machines over multiple days (as not all machines are located at the same place): * rx4640 (w/Madison) * rx2620 (w/Montecito) * rx2660 (w/Montecito) * rx6600 (w/Montvale) * rx2800 i2 (w/Tukwila) I start to test with the 3rd version and if that works on all machines, the gcc-14 snapshot it was compiled with becomes the next known-good gcc-14 snapshot. For the HP Sim platform I usually only build with the current gcc-14 snapshot (w/LRA enabled), as I can test it right after on the build host and that so far always worked. All kernels are built on a x86_64 machine using a cross-compiler built from the mentioned gcc-14 sources and with the mentioned configuration. I use "buildall" ([1]) to built this with a fixed version of binutils - 2.41 currently - to have at least one fixed point in each build. But I'll look into 2.42 soon. Another more or less fixed point is the root FS used, currently based on Debian GNU/Linux unstable. But as Debian closes shop on ia64 soon, the next test results - due in about two weeks (due to merge window) - will be done with a different root FS based on T2 ([2]). [1]: https://github.com/nathanchance/buildall [2]: https://t2sde.org/ In the following I'll send the resulting boot logs for all machines above and the HP Sim platform. Please note, AFAICT all issues (e.g. kernel oopses) present in the current logs were also already present in the past (for HP Sim for example already in 4.19.x, check the Linux stable RC and release builds for ia64 on [3]) and are **not** due to enabling LRA recently. [3]: https://github.com/johnny-mnemonic/linux-stable-rc/actions/runs/8258902207 Cheers, Frank
