2008-08-16 02:39 -!- Kirantpatil(~kiran@122.167.212.233) has joined #tux3
2008-08-16 02:39 -!- Kirantpatil(~kiran@122.167.212.233) has left #tux3
2008-08-16 11:14 -!- flips(~phillips@phunq.net) has joined #tux3
2008-08-16 11:14 -!- ChanServ changed mode/#tux3 -> -o flips
2008-08-16 14:42 -!- pgquiles(~pgquiles@132.Red-217-125-199.dynamicIP.rima-tde.net) has joined #tux3
2008-08-16 17:06 -!- bh(~billh@ip68-107-26-122.sd.sd.cox.net) has joined #tux3
2008-08-16 17:06 <bh> doesn't tux3 suck as a web server compared to apache2 ?
2008-08-16 17:17 <shapor> bh: this channel isn't about tux3 the web server
2008-08-16 17:18 -!- ChanServ changed mode/#tux3 -> +o shapor
2008-08-16 17:49 <flips> bh -> bill huery
2008-08-16 17:49 <flips> huey
2008-08-16 17:50 <shapor> ah ;)
2008-08-16 17:50 <flips> nice troll bh
2008-08-16 17:51 <flips> 
ACTION gets on a little more caffein

2008-08-16 17:53 <shapor> smelled like a troll, thats why i op'ed myself ;)
2008-08-16 17:53 <shapor> didn't make the connection 
2008-08-16 17:58 <flips> :-)
2008-08-16 17:58 <flips> bh is like that, only made one of him
2008-08-16 17:58 <flips> I am thinking... probably crazy thought
2008-08-16 17:58 <flips> but it's getting not far away from kernel port time
2008-08-16 17:59 <flips> I wonder if some of the porting could be automated with a perl script
2008-08-16 17:59 <shapor> haha
2008-08-16 17:59 <flips> see crazy above
2008-08-16 17:59 <shapor> perhaps
2008-08-16 17:59 <flips> things like changing printf to printk
2008-08-16 17:59 <flips> instead of #define printf printk
2008-08-16 18:27 <flips> a couple of big differences between the user space code and kernel: 1) kackers don't like c99 inline decls 2) bit fields... are they consistently implemented across arches? likely not
2008-08-16 18:28 <flips> could possibly isolate those differences behind inline access functions buried in header files
2008-08-16 22:06 <flips> woohoo, bitmap flush worked on the first try after plugging in the btree root
2008-08-16 22:08 <shapor> merging the changes in with my format string fix went smoothly
2008-08-16 22:08 <shapor> hg++
2008-08-16 22:37 <flips> indeed
2008-08-16 22:38 <flips> could you try it with the (unsigned int) casts removed
2008-08-16 22:38 <flips> I don't think there should be warnings but I'm prepared to be re-educated
2008-08-16 22:38 <flips> oh your merge
2008-08-16 22:38 <flips> actually I haven't tried merge in hg yet
2008-08-16 22:39 <flips> I'm glad to hear it's pleasant
2008-08-16 22:40 <flips> ok, just need to do free block and bitmap allocation is finished right up to the first kernel drop
2008-08-16 22:44 <shapor> dleaf.c:133: warning: format '%u' expects type 'unsigned int', but argument 2 has type 'long int'
2008-08-16 22:45 <shapor> just needs to be (int) to make it happy i suppose though
2008-08-16 22:45 <shapor> since we're doing pointer math we expect to be a small value
2008-08-16 23:22 <flips> I think
2008-08-16 23:23 <shapor> (int) works
2008-08-16 23:23 <shapor> although casting to int specifically for %u seems silly
2008-08-16 23:23 <flips> really
2008-08-16 23:23 <flips> it is odd that pointer difference returns long in instead of int
2008-08-16 23:23 <shapor> (unsigned int) makes more sense to me
2008-08-16 23:23 <flips> only on 64 bit
2008-08-16 23:24 <flips> we can also do the cast on the 32 bit side
2008-08-16 23:24 <shapor> and use %lu as the format string?
2008-08-16 23:25 <shapor> best way to do it is a silly thing to worry about since all the printfs are going away
2008-08-16 23:25 <flips> yes
2008-08-16 23:26 <flips> nice to keep some of the tracing around and expect it to work
2008-08-16 23:26 <shapor> well we know it will work
2008-08-16 23:26 <shapor> but i suspect printk is somewhat different anyway?
2008-08-16 23:28 <flips> very similar
2008-08-16 23:28 <flips> been honed by unix wookies for years
2008-08-16 23:32 <flips> ok, put back pretty much the way you had it
2008-08-16 23:32 <flips> yuck huh?
2008-08-16 23:33 <flips> 
ACTION done for today

2008-08-16 23:34 <shapor> hm
2008-08-16 23:34 <shapor> have you thought about the user interface to snapshot data
2008-08-16 23:35 <shapor> say i want to copy a directory and all the previous version of it
2008-08-16 23:35 <flips> snapshot data or snapshot?
2008-08-16 23:35 <shapor> to say, another tux3 filesystem
2008-08-16 23:35 <flips> all previous versions, no
2008-08-16 23:35 <flips> just a specified version
2008-08-16 23:35 <flips> copy with history is an interesting challenge
2008-08-16 23:36 <shapor> could be useful for quite a few things
2008-08-16 23:36 <flips> sounds hard
2008-08-16 23:36 <shapor> i know, thats why i asked
2008-08-16 23:36 <flips> but if the use case is compelling...
2008-08-16 23:37 <flips> copy history does sounds useful
2008-08-16 23:37 <shapor> very
2008-08-16 23:37 <shapor> say you take daily snapshots
2008-08-16 23:37 <shapor> but you want to do a weekly backup to tape
2008-08-16 23:37 <shapor> which includes all those incremental daily changes
2008-08-16 23:37 <shapor> or even hourly
2008-08-16 23:38 <flips> easy, back up a bunch of deltas
2008-08-16 23:38 <shapor> can you get a delta for a particular directory?
2008-08-16 23:38 <flips> that is the plan
2008-08-16 23:38 <flips> of a given file
2008-08-16 23:38 <shapor> what is the interface?
2008-08-16 23:38 <flips> some ddlink thing
2008-08-16 23:38 <flips> with some c program driving the ddlink
2008-08-16 23:39 <flips> in other words, go crazy
2008-08-16 23:39 <shapor> yeah
2008-08-16 23:39 <shapor> but that wasn't the usecase i originally thought of
2008-08-16 23:39 <shapor> i was trying to simplify it
2008-08-16 23:39 <flips> probably best to think about the primitive ops and think how to link them up to do complex things
2008-08-16 23:40 <flips> thinking about an organized delta store
2008-08-16 23:40 <flips> that you can not only write to but read from... and do what with?
2008-08-16 23:40 <flips> something like the guy did with zumastor maybe
2008-08-16 23:41 <flips> mountable deltas
2008-08-16 23:41 <flips> that would rule the universe
2008-08-16 23:41 <flips> but hard, probably
2008-08-16 23:41 <shapor> yeah migrate them to nearline storage
2008-08-16 23:41 <shapor> i bet you think doing it with ddsnap delta files was hard too though
2008-08-16 23:41 <shapor> and someone did it ;)
2008-08-16 23:42 <shapor> speaking of whom, you should loop him in about tux3
2008-08-16 23:44 <flips> yes
2008-08-16 23:44 <flips> remember the subject line?
2008-08-16 23:45 <shapor> ddloop perhaps?
2008-08-16 23:45 <flips> I thought that was my name
2008-08-16 23:46 <shapor> "Backups using ddsnap
2008-08-16 23:46 <flips> http://www.nomorevoid.com/downloads/dm-ddloop.tar.gz
2008-08-16 23:47 <shapor> "
2008-08-16 23:47 <flips> right on our home page
2008-08-16 23:47 <shapor> was the subject on the list
2008-08-16 23:47 <shapor> yeah
2008-08-16 23:47 <flips> got it
2008-08-16 23:48 <shapor> why dont google groups archives appear in google search results?
2008-08-16 23:48 <flips> because goog suxorx?
2008-08-16 23:48 <shapor> http://groups.google.com/group/zumastor/browse_thread/thread/c95970acdc2e31ca/cc9931d18043f31f
2008-08-16 23:49 <shapor> http://www.google.com/search?q=%22backups+using+ddsnap%22
2008-08-16 23:49 <shapor> doesn't find it unfortunately
2008-08-16 23:57 <flips> googlegroups is really too lame to have a mailing list on
2008-08-16 23:57 <shapor> it would be fine is gmane was subscribed
2008-08-16 23:57 <flips> I'll set up a new zumastor mailing list
2008-08-16 23:59 <flips> t      A following integer conversion corresponds to a ptrdiff_t argument -- printf man page