Hi, Phil Sutter wrote,
Hi Waldemar,
On Tue, Jan 07, 2025 at 10:57:00AM +0100, Waldemar Brodkorb wrote:
Hi Phil, Waldemar Brodkorb wrote,
Hi Phil, Phil Sutter wrote,
Without it, on these architectures 'lvm pvscan' will busy-loop printing:
| io_getevents failed: No such file or directory
What C library you are using? Nowadays glibc/uClibc-ng and musl have 64 Bit time_t on all supported 32 Bit architectures with a recent enough kernel (=>5.1.0).
I'm using uClibc-ng version 1.0.50. B-)
I think using this kernel symbol is only required for C libraries not having 64 Bit time_t. So we could remove the symbol entirely in OpenADK.
While your point is valid regarding 64bit time_t, the problem here seems to be io_getevents syscall which expects a 32bit time_t in parameters. I guess this can't change due for compatibility reasons and one had to introduce a new syscall which accepts the larger time_t. Or is this just covered by the libc wrappers?
Good question, I don't know :-/
I checked this on my Cubox-i (ARM 32 Bit) and your patch is valid. It seems this kernel symbol is even required for time64 enabled userland.
Thanks for validating!
BTW: Does gcc-14 work for you? It built fine for me but I got compile errors building packages. Went with gcc-12 for now.
I use the default gcc-13 for now, which compiles all packages I use fine. Yes, gcc-14 will break a lot of packages.
Do you have any issue with gcc-13?
best regards Waldemar