Home
release-team@conference.openafs.org
Friday, October 4, 2019< ^ >
Room Configuration
Room Occupants

GMT+0
[12:58:43] meffie joins the room
[15:02:10] <kaduk@jabber.openafs.org/barnowl> Hmm, am I secretly disconnected?
[15:02:38] <kaduk@jabber.openafs.org/barnowl> (signs point to 'no')
[15:05:22] <meffie> good morning.
[15:05:52] <meffie> sorry, lost track of time.
[15:06:01] <kaduk@jabber.openafs.org/barnowl> It's friday; understandable :)
[15:06:28] <meffie> heh
[15:06:43] wiesand joins the room
[15:07:07] <kaduk@jabber.openafs.org/barnowl> While we see if anyone else is coming, I left aquestion for you on
13883
[15:07:10] <wiesand> sorry, was late anyway, then connecting took some time
[15:11:22] <wiesand> re 1.8.4: any showstoppers ore late "one more we must have"s?
[15:11:33] <kaduk@jabber.openafs.org/barnowl> not from here
[15:12:08] <meffie> none here
[15:12:17] <wiesand> if not, I'd merge the "NEWS" and "version strings" changes now and ask ben for the tag
[15:12:33] <kaduk@jabber.openafs.org/barnowl> sounds good
[15:12:49] <meffie> thank you
[15:13:53] <wiesand> [working on that]
[15:18:03] <wiesand> changes arrived in git - Ben, could you please tag 501dbbe4e2fbcb8e71af9f695e8e34911c44d581 as openafs-stable-1_8_4 ?
[15:18:57] <kaduk@jabber.openafs.org/barnowl> working on it
[15:19:11] <wiesand> from there I could work on getting 1.8.4 out at my pace
[15:20:00] <kaduk@jabber.openafs.org/barnowl> * [new tag]               openafs-stable-1_8_4 -> openafs-stable-1_8_4
[15:20:36] <wiesand> (things are a bit chaotic here… mother is in hospital since today - nothing too serious as it seems, but lots of phone calls etc.)
[15:21:13] <kaduk@jabber.openafs.org/barnowl> Yikes!  I hope it remains nothing too serious
[15:21:18] <meffie> oh, hope she will be ok.
[15:22:10] <wiesand> thanks
[15:23:17] <wiesand> chances are it will work out well this time - I'm glad she's actually there and didn't accept the offer to stay at home over th weekend at her own responsibility
[15:23:58] <wiesand> thanks for the tag Ben
[15:24:55] <kaduk@jabber.openafs.org/barnowl> no problem
[15:25:03] <wiesand> looks like 1.8.4 is on track then
[15:25:54] <wiesand> unless there are any vetos, I'd not wait very long for reviews of the web change this time
[15:28:52] <wiesand> I'll try to get it done and pushed out over the weekend and announced on Monday
[15:29:21] <kaduk@jabber.openafs.org/barnowl> yay
[15:29:25] <wiesand> And then on to 1.8.5pre1...
[15:29:51] <wiesand> no progress on 1.6.24 here, sorry
[15:30:05] <wiesand> on to master/1.9.0?
[15:30:37] <kaduk@jabber.openafs.org/barnowl> Not a whole lot of progress there, though now that 1.8.4 is tagged we
can update the 1.9.0 NEWS and push that
[15:30:53] <meffie> i'll check the list of gerrits for 1.8.5 (after this meeting)
[15:31:01] <kaduk@jabber.openafs.org/barnowl> I did rebase the first pthread-bos commit ... to discover that I
rebased past the softsig/windows changes and will need to tweak it.
[15:31:18] <meffie> thank you for the rebase!
[15:31:46] <wiesand> pretty long list already (which is both good and bad ;-)
[15:32:27] <meffie> some of the items have been in master for a while, i found them while working on 1.9.0 NEWS
[15:33:00] <meffie> seems 1.8.5 was the next window to pull them up.
[15:36:22] <kaduk@jabber.openafs.org/barnowl> If we're still going for master topics, maybe we can talk about the
network-address-formatting topic.
[15:36:35] <meffie> ok
[15:36:56] <kaduk@jabber.openafs.org/barnowl> Are people converging on doing our own printf formatter?  I confess
the idea does not fill me with a huge amount of excitement, though it
may well end up being the right thing to do.
[15:37:48] <kaduk@jabber.openafs.org/barnowl> (discussion in 13832)
[15:38:14] <meffie> yes, i assumed we would not be excited about a custom printf formatter, but maybe that would be better in the long run.
[15:38:58] <meffie> i thing it could be nice to do something like the linux printf magic %p<code> extensions, i think apache apr does the same.
[15:39:10] <meffie> s/thing/think/
[15:39:18] <kaduk@jabber.openafs.org/barnowl> My personal preference/bias is to investigate something "sbuf-like"
more before concluding that printf formatting is best
[15:40:41] <meffie> hmm, that seems like a lot of code. isn't this for log message formatting?
[15:41:09] <kaduk@jabber.openafs.org/barnowl> That (sbuf-like) would allow for custom printf formatting without
requiring all callers in the tree to convert over to it
[15:42:36] <kaduk@jabber.openafs.org/barnowl> The immediate use is log message formatting, yes.  We don't need all
of sbuf, at least not at first, and can customize what we have to suit
our needs.  I guess it's worth thinking about whether we will need to
prepare structured output, string or otherwise, for much other than
log messages.
[15:42:57] <kaduk@jabber.openafs.org/barnowl> Given that we already use XDR for the protocol RPCs, maybe we're just
going to be doing logging stuff.
[15:43:31] <meffie> sbuf is something different that the example...
FOO *foo = FOO_new();
FOO_append_string("some static data ");
...
[15:43:37] <meffie> ?
[15:44:48] <kaduk@jabber.openafs.org/barnowl> It's similar, but presumably different, yes, since we'd customize
something that meets our needs.
[15:45:01] <meffie> ok
[15:45:34] <meffie> we couldnt do something like?
[15:46:21] <meffie> ViceLog((1, "the host %pH is borked.", host));
[15:47:07] <kaduk@jabber.openafs.org/barnowl> We could probably do that (ViceLog), but I think it's pretty messy to
try to only handle certain % sequences, and you end up reimplementing
all of printf
[15:47:08] <meffie> and register a formatter function at startup ?
[15:47:58] <kaduk@jabber.openafs.org/barnowl> Which, I guess maybe you end up reimplementing all of printf anyways
for some of the other options.
[15:48:08] <kaduk@jabber.openafs.org/barnowl> Or do I misunderstand how this would work?
[15:48:23] <meffie> yes, we'd have to implement all of printf to do that. i assume we could just reuse a bsd or apr formmatter, or the roken xyzprint?
[15:49:46] <meffie> we could have opr_formatter() that implements printf for logging messages
[15:49:47] <kaduk@jabber.openafs.org/barnowl> Hmm, roken would be the default option, but then there's the dilemma
of whether to preserve upstream roken for ease of future imports or to
modify it...
[15:52:52] <meffie> that's a good question.
[15:53:11] <meffie> i feel if we are changing something we need something in tree.
[15:53:45] <meffie> i'm not sure if it is possible to get roken to support printf extensions.
[15:54:12] <meffie> here's the apache version (as a data point): https://github.com/apache/apr/blob/trunk/strings/apr_snprintf.c
[15:54:48] <kaduk@jabber.openafs.org/barnowl> (Spot checking FreeBSD, I see char ip4buf[INET_ADDRSTRLEN]; and
sbuf_printf("%s...", inet_ntoa_r(., ip4buf)), so they don't have super
magic already)
[15:56:18] <meffie> for our use, we would need to change all the logging in host.c, and it would be really messy.
[15:57:00] <meffie> but possible (mechanical)
[15:57:09] <kaduk@jabber.openafs.org/barnowl> Er, which change would be really messy?  Switching to something with
not-a-printf-format-code interface?
[15:57:39] <meffie> yes, that's what i mean. there's a lot of printf formats.
[15:59:01] <meffie> well, lot is relative.
[16:01:01] <kaduk@jabber.openafs.org/barnowl> My intuition is telling me that there's a way to keep the mess
contained within the ViceLog implementation, whether it's
auto-declaring storage and doing the inet_ntoa_r calls or
tracking/freeing dynamically allocated buffers or something else.
[16:01:19] <meffie> so the options are: 1. helper functions (like PrintInode() and the new ones in gerrit), 2. convert to sbuf-like functions. 3. implement our own formatter with magic %p<codes> ?
[16:02:30] <meffie> ah 4. new logging api, maybe like the fstrace functions? tagname, values...?
[16:04:03] <meffie> more like: afs_Trace2(afs_iclSetp, CM_TRACE_FLUSHV, ICL_TYPE_POINTER, avc, ICL_TYPE_INT32, avc->f.states);
[16:04:31] <meffie> and let the logger use something like sbuf_ calls to build the message.
[16:05:38] <meffie> and we could have "tags" for our types that need to be formatted?
[16:06:06] <kaduk@jabber.openafs.org/barnowl> It sounds like we have some more pondering to do.
[16:06:13] <meffie> :)
[16:06:15] <meffie> yes
[16:06:22] <kaduk@jabber.openafs.org/barnowl> But we're past the hour, so maybe we should ask if there are other
topics for today?
[16:06:51] <meffie> yes, thanks for the discussion and ideas.
[16:07:19] <wiesand> Nothing else from me for today, sadly
[16:07:57] <kaduk@jabber.openafs.org/barnowl> Okay.  Thanks everyone!
[16:08:27] <wiesand> Let's adjourn then. Thanks a lot Ben & Mike!
[16:08:34] <meffie> thanks! have good weekend.
[16:09:02] <kaduk@jabber.openafs.org/barnowl> you, too!
[16:10:10] wiesand leaves the room
[16:32:42] meffie leaves the room
[19:11:51] meffie joins the room
[19:11:57] meffie leaves the room
[23:57:48] mvita leaves the room