14:30:25 <gitomat> [illumos-gate] 16625 audio: suggest parentheses around assignment used as truth value -- Toomas Soome <tsoome⊙mc>
20:10:54 <gitomat> [illumos-gate] 16630 libsecdb: i.rbac is assuming C locale -- Toomas Soome <tsoome⊙mc>
20:25:50 <jbk> rmustacc: currently we parse the NVMe controler 'model' field using sata_split_model() to try to extract the vendor, though that's assuming the vendor is the first 'word'
20:26:09 <jbk> this doesn't seem to be the case for at least this Toshiba NVMe we have...
20:26:50 <jbk> i'm wondering if it might be better to somehow leverage the vid instead of trying to parse the model string...
20:26:54 <jbk> (or something else)
20:27:19 <jbk> (so the vendor shows up as the fallback 'NVMe')
21:39:57 <gitomat> [illumos-gate] 16632 unix: conflicting types for 'cpuid_getchiprev' -- Toomas Soome <tsoome⊙mc>
23:39:17 <tozhu> does illumos support posix_fadvise api? I hit issue in postgresql that set effective_io_concurrency=2, then it report the platform can’t support posix_fadvise , (I use zfs on illumos), so I just want to comfirm if it is true illumos does not support posix_fadvise on zfs
23:39:22 <tozhu> best wishes
23:45:57 <rmustacc> jbk: In general I think the stuff we fake for block devices is kind of odd imho.
23:46:18 <rmustacc> The NVMe spec doesn't have a string name for vendors just the PCI vid in the identify controller bit.
23:47:12 <rmustacc> If you passed through the vid, it'd have to be userland that did the translation into something there. We don't really want a kernel version of libpcidb.
23:49:31 <rmustacc> tozhu: It's not wired up, but shouldn't error.