gistfile1.txt

2017-11-15 01:01:21	gchristensen	*braces to ask a painful question* are there any known patches to add back support for the 2.6.32 kernel for glibc 2.26?
2017-11-15 01:02:05	gchristensen	I'm marooned on RHEL 6 and have been using Nix to provide more recent package sets, including a more recent glibc. they upgraded to 2.26 recently, which means I'm now stuck back in time for Nix too.
2017-11-15 01:22:56	azanella	gchristensen, nops, you will need to check which was assumed to be removed with minimum 3.2 and backported
2017-11-15 01:23:38	azanella	from Joseph message from '139ace95756a6' it assumes presence of getcpu (x86_64), recvmmsg (not always through its own syscall, sometimes only through socketcall), sendmmsg (likewise), /proc/$pid/task/$tid/comm, f_flags from statfs, prlimit64.
2017-11-15 01:24:16	azanella	also I added a lot of consolidation implementations that changed slight which syscall was used, it might not be supported on 2.6.32
2017-11-15 01:24:33	gchristensen	I understand
2017-11-15 01:25:02	azanella	it is a matter to adjust to enable 2.6.32 and check which is requires adjustments
2017-11-15 01:25:30	gchristensen	yeah, I'm pretty sure that'll be more than I can handle, especially if it was deemed worth dropping support officially
2017-11-15 01:26:06	gchristensen	just unfortunate for some minor use cases like mine, but I can appreciate how much burden old software like this places on projects
2017-11-15 01:26:50	gchristensen	thank you for the information, azanella 
2017-11-15 01:26:54	azanella	yw
2017-11-15 01:27:42	azanella	gchristensen, it thinks it is doable, although it would require some work and the work is mostly 'reverting' stuff
2017-11-15 01:28:45	gchristensen	ok, I'll keep that in mind
2017-11-15 01:29:06	gchristensen	I'm going to first try the angle of let's upgrade to a kernel from this decade 

• getcpu, *mmsg, fflags in statfs
• prlimit64 is also mentioned in the sources
• verify /proc/$pid/task/$tid/comm on a running instance
• understand

  also I added a lot of consolidation implementations that changed slight which syscall was used, it might not be supported on 2.6.32

• find a reasonable way to disable the check. We would probably prefer to still get a readable error on truly unsupported kernels. (try asking glibc upstream first?)