06:25:55 <***> Buffer Playback...
06:25:55 <moneromooo> [14:16:33] I see. Should be simple enough to test whether you're writing on... what, a device ?
06:25:55 <hyc> [14:22:37] yeah. in particular a block device
06:25:55 <hyc> [14:23:07] I had support for character devices too but that doesn't perform well
06:25:55 <hyc> [14:25:09] at least in Linux, block devices have a write-thru cache. character devices appear to have no cache at all, so read perf sucks
06:25:55 <hyc> [14:27:00] but still I'm wondering why bother with those startup checks at all?
06:25:55 <moneromooo> [14:27:20] People were confused by the late error IIRC.
06:25:55 <hyc> [14:27:29] it's not even checking on WIN32 and generally windows users are the ones who know the least about their computers
06:25:55 <moneromooo> [14:28:30] If it causes more trouble than it helps now, feel free to remove tbh.
06:25:55 <moneromooo> [14:29:07] And maybe rig the late error to add, eg, " (maybe your fs doesn't mmap)"
06:25:55 <moneromooo> [14:29:26] If you can distinguish between errors at that point.
06:25:55 <hyc> [14:29:51] hm... I'll take a look. at the moment I think all you'll get is mdb_env_open() returns EINVAL
06:25:55 <hyc> [14:30:09] what FS did someone use that didn't support mmap?
06:25:55 <hyc> [14:31:02] even NFS claims to support it, however badly
06:25:55 <moneromooo> [14:31:53] Some FAT on USB drives I think.
06:25:55 <hyc> [14:32:20] ok I'll try that
06:25:55 <moneromooo> [14:32:21] NFS does ring a bell actually. Maybe grep "mmap nfs" on closed github issues ?
06:25:55 <hyc> [14:32:47] well. NFS won't refuse the request. it will just perform poorly
06:25:55 <hyc> [14:34:36] fat32 USB flash drive works fine here
06:25:55 <hyc> [14:36:02] I wonder if it was exfat or something
06:25:55 <hyc> [14:41:21] nope. exfat working too
06:25:55 <hyc> [14:42:26] normal fat32 wouldn't work since it has a 4GB filesize limit
06:25:55 <hyc> [14:42:39] I mean, it would work until it hit that size limit
06:25:55 <mj-xmr> [04:22:37] selsta, are you sure, that 7769 and 7770 shouldn't be kept as separate PRs?
06:25:55 <mj-xmr> [04:23:18] in 7770 it's strictly about Boost, while in 7769 it's about Ubuntu 18.04 with everything else defaulted.
06:25:55 <mj-xmr> [04:29:36] Also, don't we want to test the minimum advertised Boost version in 7770? Without this test, we'd still think that 1.58 is the minimal version. The same thing can happen to 1.62 in the future, so it'd be good to have it tested. Ccache handles it.
06:25:55 <***> Playback Complete.