Perl 6 - the future is here, just unevenly distributed

IRC log for #gluster-dev, 2014-11-04

| Channels | #gluster-dev index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:46 shyam joined #gluster-dev
01:58 shyam joined #gluster-dev
02:57 bala joined #gluster-dev
03:01 bala joined #gluster-dev
03:05 anoopcs joined #gluster-dev
03:15 ryao__ joined #gluster-dev
03:40 shubhendu__ joined #gluster-dev
03:41 kanagaraj joined #gluster-dev
04:11 atalur joined #gluster-dev
04:22 rafi1 joined #gluster-dev
04:22 Rafi_kc joined #gluster-dev
04:31 jiffin joined #gluster-dev
04:33 anoopcs joined #gluster-dev
04:34 bharata-rao joined #gluster-dev
04:36 jiffin joined #gluster-dev
04:40 nkhare joined #gluster-dev
04:41 lalatenduM joined #gluster-dev
04:42 anoopcs joined #gluster-dev
04:44 hagarth joined #gluster-dev
04:48 ndarshan joined #gluster-dev
04:48 atinmu joined #gluster-dev
04:49 nishanth joined #gluster-dev
04:56 kdhananjay joined #gluster-dev
04:58 soumya joined #gluster-dev
05:13 ppai joined #gluster-dev
05:42 pranithk joined #gluster-dev
05:42 nishanth joined #gluster-dev
05:45 ndarshan joined #gluster-dev
05:46 shubhendu__ joined #gluster-dev
06:22 anoopcs joined #gluster-dev
06:23 soumya joined #gluster-dev
06:26 atalur joined #gluster-dev
06:26 ppai joined #gluster-dev
06:36 davemc joined #gluster-dev
06:40 soumya joined #gluster-dev
06:50 itisravi joined #gluster-dev
06:57 shubhendu__ joined #gluster-dev
06:57 ndarshan joined #gluster-dev
06:59 nishanth joined #gluster-dev
07:07 JustinClift joined #gluster-dev
07:07 ndk joined #gluster-dev
07:08 kkeithley joined #gluster-dev
07:09 rgustafs joined #gluster-dev
07:13 atalur joined #gluster-dev
07:16 Humble hagarth++ thanks!
07:16 glusterbot Humble: hagarth's karma is now 19
07:17 atinmu joined #gluster-dev
07:18 kkeithley joined #gluster-dev
07:20 bala joined #gluster-dev
07:20 anoopcs joined #gluster-dev
07:46 bala joined #gluster-dev
07:48 atinmu joined #gluster-dev
08:04 purpleidea *** testing travis ***
08:08 hagarth xavih: ping, is there a version of datalab.es in English? trying to link to your site from a blog post :)
08:14 atalur joined #gluster-dev
08:42 bala joined #gluster-dev
08:43 ndevos hey lalatenduM! any progress on the 3.5.3beta2 packages?
08:44 davemc joined #gluster-dev
08:46 lalatenduM ndevos, working on it
08:46 ndevos lalatenduM: okay, thanks!
08:47 ndevos lalatenduM: should I already move the bugs to ON_QA and post on my blog?
08:47 lalatenduM ndevos, not now may be some time later, will let you know
08:48 ndevos lalatenduM++ works for me :)
08:48 glusterbot ndevos: lalatenduM's karma is now 34
08:48 lalatenduM :)
08:49 vikumar joined #gluster-dev
08:50 atinmu joined #gluster-dev
09:04 hagarth joined #gluster-dev
09:16 rgustafs joined #gluster-dev
09:32 xavih hagarth: we don't have an english version of www.datalab.es (we'll work on it), but you can use dlnas.com. It has an english version and contains more updated information.
09:36 ppai joined #gluster-dev
09:36 soumya joined #gluster-dev
09:36 nishanth joined #gluster-dev
09:36 ndk joined #gluster-dev
09:36 JustinClift joined #gluster-dev
09:36 kkeithley joined #gluster-dev
09:36 anoopcs joined #gluster-dev
09:36 atalur joined #gluster-dev
09:36 vikumar joined #gluster-dev
09:36 atinmu joined #gluster-dev
09:37 hagarth joined #gluster-dev
09:38 davemc joined #gluster-dev
09:41 shubhendu__ joined #gluster-dev
09:47 ndarshan joined #gluster-dev
09:51 nishanth joined #gluster-dev
09:53 atinmu joined #gluster-dev
10:25 lalatenduM ndevos, somehow I have missed/deleted your 3.5.3beta3 release mail, can you plz forward again
10:30 davemc joined #gluster-dev
10:31 ndevos lalatenduM: fwd'd
10:35 hagarth joined #gluster-dev
10:35 hagarth xavih++ thanks for that :)
10:35 glusterbot hagarth: xavih's karma is now 7
10:41 pranithk left #gluster-dev
10:42 lalatenduM ndevos, thanks
10:52 lalatenduM ndevos, the previous beta for 3.5.3 was glusterfs-3.5.3-0.1.beta1 and my beta2 I am building glusterfs-3.5.3-0.1.beta2, is it ok
10:52 lalatenduM ndevos, usually we bump up 0.1.beta1 0.2.beta2, but I forgot
10:53 ndevos lalatenduM: oh, that does not matter 0.1.beta2 works fine
10:55 lalatenduM ndevos, cool
10:56 lalatenduM ndevos, however I did not get ur fwd'd  mail
10:58 hagarth ndevos: the bug triage meeting is in one hour right?
10:58 hagarth ndevos: as far as my AI goes, I am still in the process of identifying one such person to help our efforts.
10:59 atinmu joined #gluster-dev
10:59 ndevos hagarth: yes, in one hour - thanks for the update :)
11:00 ndevos lalatenduM: no? to your @redhat.com address?
11:01 ndevos lalatenduM: subject="glusterfs-3.5.3beta2 released"
11:01 lalatenduM ndevos, nope, what is the subj line, thr might be something wrong with my thunderbird
11:01 hagarth ndevos: my calendar still shows it for 2 hours from now :|
11:02 ndevos lalatenduM: that email was not sent to the list, so maybe it's somewhere in a different folder/filter?
11:03 ndevos hagarth: hmm, I can update the zimbra invite, or delete it - I get many "tentative" and "declined" messages from people on the lists that delete the entry  in their calendar
11:03 lalatenduM ndevos, the search found it. but not sure why it is coming as per date sort in inbox
11:04 ndarshan joined #gluster-dev
11:05 nishanth joined #gluster-dev
11:05 shubhendu__ joined #gluster-dev
11:07 hagarth ndevos: true, not sure what is the ideal way out here
11:07 ndevos hagarth: I'm not sure either... and it seems that zimbra had the meeting incorrect in my calendar too?!
11:08 * ndevos just deleted the invite, and will need to think about re-scheduling one
11:10 ndevos hagarth: Manu asks on -devel to revert http://review.gluster.org/9016 because it causes all regression tests to fail
11:11 ndevos hagarth: I can hit the "revert change" button, if you want me to?
11:18 kkeithley1 joined #gluster-dev
11:18 ndevos revert posted as http://review.gluster.org/9040
11:30 lalatenduM joined #gluster-dev
11:34 edward1 joined #gluster-dev
11:40 atinmu joined #gluster-dev
11:42 spandit joined #gluster-dev
11:44 raghu` joined #gluster-dev
11:53 atalur joined #gluster-dev
11:56 soumya_ joined #gluster-dev
11:57 ndevos REMINDER: Gluster Bug Triage meeting starting in a few minutes in #gluster-meeting
12:08 jdarcy joined #gluster-dev
12:28 hagarth joined #gluster-dev
12:30 badone_ joined #gluster-dev
12:34 raghu` left #gluster-dev
13:07 bala joined #gluster-dev
13:13 ndevos kkeithley_: on libgfapi.so - could VERS_ in the .map.in be replaced by GFAPI_ so that the version looks a little nicer?
13:14 ndevos kkeithley_: I'm building packages now, and want to test update/installation on RHEL-6.6 later today
13:15 ndevos kkeithley_: also, prefixing each symbol with an 'x' looks a little awkward, how about a '_' as prefix, or some _PUB postfix?
13:16 ndevos the versioned symbols themselved look fine with objdump, but maybe we can make things more obvious/easier/less-confusing for libgfapi developers
13:17 * ndevos goes for lunch, will be back later!
13:36 kkeithley_ other than leading _ (or __) is reserved,
13:40 lalatenduM kkeithley, ndevos Humble you guys have some time , I have some questions on 3.5.3beta2
13:40 kkeithley_ Not that others haven't already used __, but I don't like perpetuating or reinforcing bad practices.
13:40 kkeithley_ fire away
13:41 hagarth merging the reverted patch now
13:41 lalatenduM kkeithley, I committed spec file changes for 3.5.3bete2 on f21 branch as rawhide had 3.6.0 changes as of now
13:41 lalatenduM then for f20 branch , I rebased it from f21 branch
13:42 kkeithley_ that's right. Or that's what I'd have done too
13:42 lalatenduM for f19 , I tried doing the same , but it has lots of conflicts , so I aborted the rebase , what should I do now
13:44 lalatenduM I can commit the same changes in f19 by taking spec file from f21 branch and it looks easy than a rebase
13:44 kkeithley_ Hmm.  that's odd
13:44 kkeithley_ odd that you had lots of conflicts
13:45 kkeithley_ yes, you could just copy the spec from the f20 or f21 branch into the f19 branch. That should be oka
13:45 kkeithley_ the result is the same
13:45 kkeithley_ should be the same
13:47 lalatenduM kkeithley, may be I missed something in the rebase f19, do you want to give it a try
14:00 ira joined #gluster-dev
14:10 hchiramm joined #gluster-dev
14:12 kkeithley_ did you already commit and push the changes to the other branches?
14:15 kkeithley_ nm, no you didn't.
14:15 kkeithley_ send me the spec?
14:15 bala1 joined #gluster-dev
14:16 soumya_ joined #gluster-dev
14:25 lalatenduM kkeithley, yes, I have committed changes to f21 and f20
14:26 lalatenduM kkeithley, you need to following , 1. fedpkg switch-branch f19 2. git rebase f21
14:27 lalatenduM kkeithley, I again tried the rebase , there are multiple conflicts
14:27 lalatenduM I opted git checkout --theirs glusterfs.spec
14:28 lalatenduM + git add glusterfs.spec + git rebase --continue
14:28 kkeithley_ I don't see email on fedora-updates about a commit
14:28 kkeithley_ And I just did another pull and I'm not getting your spec
14:28 lalatenduM kkeithley, there was a email for f21
14:29 kkeithley_ sorry, my mistake about the emails
14:30 kkeithley_ gah, I'm hating git, or git is hating me.
14:30 kkeithley_ oh, wait
14:31 kkeithley_ there it is
14:33 kkeithley_ I get merge conflicts as well
14:34 lalatenduM ohh :(
14:34 lalatenduM what did I miss
14:35 lalatenduM you mean got merge conflicts in f20 and f21 branches as well
14:35 kkeithley_ no, just trying to merge origin/f20 into f19 branch
14:36 lalatenduM ok
14:36 kkeithley_ just copy the spec from f20 branch into f19
14:36 shyam joined #gluster-dev
14:37 lalatenduM kkeithley, yeah that is better , on it
14:37 lalatenduM kkeithley++
14:37 glusterbot lalatenduM: kkeithley's karma is now 32
14:41 lalatenduM kkeithley, I think I need to add the sources file too, isn't it?
14:45 lalatenduM kkeithley, done
14:55 lalatenduM kkeithley, has send you a mail, let know what do you think
14:57 nkhare joined #gluster-dev
14:59 ndevos kkeithley_: _ and __ are reserved when they are used as pre- *and* postfix, but we use __ as prefix for "use inside a lock"
15:00 ndevos kkeithley_: but, yes, I also think it isn't too pretty, _PUB, or maybe LIB_ as post/prefix?
15:03 kkeithley_ I don't want to keep doing bad things just because we've done them before. ;-)
15:03 kkeithley_ right now I'm using pub_ (and priv_) prefix. But it could just as easily be uppercase
15:04 kkeithley_ what about Anders' comments, any thoughts?
15:04 ndevos no, I agree, but lets not introduce new bad things like a 'x' prefix - unless that 'x' means something :)
15:04 kkeithley_ will commit in a few minutes
15:04 ndevos uh, duplicate? i've not noticed that, lets see
15:05 ira joined #gluster-dev
15:06 ndevos kkeithley_: not duplicates, but there is this strange double @@, like glfs_creat@@VERS_3.6.0
15:07 ndevos seems to be for all(?) 3.6.0 symbols
15:07 vikumar joined #gluster-dev
15:07 kkeithley_ @@ is the "default" See Ulrich Drepper's paper on writing shared libs (on intel web site)
15:08 kkeithley_ https://software.intel.com/sites/d​efault/files/m/a/1/e/dsohowto.pdf
15:09 kkeithley_ i.e. which version to link with by default
15:11 kkeithley_ lalatenduM: I don't know what to tell you about the merge in the kk-el5 branch
15:12 lalatenduM kkeithley, should I revert it? and can just go with adding specfile and sources file and not a rebase
15:12 kkeithley_ I believe that'll work, so yes
15:12 ndevos kkeithley_: ah, okay, thanks for the pointer!
15:14 lalatenduM kkeithley, I am going for a "git revert HEAD"
15:14 lalatenduM as the HEAD is my lat commit
15:14 lalatenduM last*
15:15 kkeithley_ I just copied the style from that. I'm going to try his typeof suggestion. Not sure what to say about all the dupe symbols for every version, other than this is a quick "get it out the door" fix. Hopefully we don't make any mistakes that box us into a corner later
15:18 kkeithley_ Or not do the typeof suggestion. I should have written macros for this
15:21 jobewan joined #gluster-dev
15:25 ndevos kkeithley_: oh, yes, and I agree that the duplicate symbols could be better - we would only need duplicates for symbols that got backported, I think
15:39 kkeithley_ well, we don't know which ones those are going to be yet
15:40 ndevos yeah, and I think we can drop older symbol-versions later, they did not exist yet, so nothing linked against them
15:41 bala joined #gluster-dev
15:42 lalatenduM kkeithley++ for http://review.gluster.org/#/c/9036
15:42 glusterbot lalatenduM: kkeithley's karma is now 33
15:42 lalatenduM ndevos++ too :)
15:42 glusterbot lalatenduM: ndevos's karma is now 42
15:46 kkeithley_ ugh, premature karma
15:46 hagarth ndevos++
15:46 glusterbot hagarth: ndevos's karma is now 43
15:46 ndevos yeah, that too, and it should probably be more of a kkeithley++ anyway ;)
15:46 glusterbot ndevos: kkeithley's karma is now 34
15:47 kkeithley_ so you think for now we just need to set symbol version of the version it first appeared in.
15:47 kkeithley_ i.e. the asm line
15:47 kkeithley_ and no aliases until we forward part an old version?
15:49 ndevos I think the version of the symbol should be the version where is occured first
15:49 kkeithley_ agreed
15:50 ndevos when I look with 'readelf -s' at libvirt.so, most symbols have only one version - newer versions likely changed the functin signature
15:51 kkeithley_ yes... (bad idea IMO to change function signatures. better to add a new function with a different name. That's what my experience at the X Consortium taught me anyway.)
15:53 ndevos I agree, a different name makes it much more developer friendly too
15:54 kkeithley_ I'm going to drop all the aliases (duplicate symbols), keep one version of each function w/ version it appeared in.
15:55 kkeithley_ one version of each symbol+version using the version it appeared in
16:00 wushudoin joined #gluster-dev
16:02 ndevos cool
16:02 kkeithley_ do you think that's what Anders is saying?
16:03 kkeithley_ and do we think that's the correct thing to do?
16:04 kkeithley_ from a "mechanical" standpoint, adding symbol versions is "easy." It's the semantics that's hard.
16:06 ndevos yeah, I think that is what anders means
16:06 ndevos and, how about changing the VERS_ into GFAPI_ or something?
16:07 kkeithley_ I did that already in change 10
16:07 kkeithley_ change 11 coming in a bit
16:09 ndevos oh, lol, I cant keep up!
16:10 kkeithley_ It's a work-in-progress ;-)
16:11 ndevos obviously
16:16 lalatenduM ndevos, builds are done for 3.5.3beta2
16:17 lalatenduM ndevos, now we need to put it on d.g.o
16:17 ndevos lalatenduM++ awesome! many thanks
16:18 glusterbot ndevos: lalatenduM's karma is now 35
16:21 lalatenduM kkeithley, I think RPMs for glusterfs-3.4.6beta2 is pending?
16:22 kkeithley_ pending, from me?
16:22 lalatenduM kkeithley, nope, from us me and Humble :)
16:22 kkeithley_ ah, okay. good, and thanks
16:38 jiffin joined #gluster-dev
16:47 hagarth joined #gluster-dev
16:50 kkeithley_ there's patch set 11 of symbol versions. take a look
16:50 kkeithley_ please
16:50 ndevos :)
16:55 kkeithley_ alstublieft!  hah
16:57 ndevos :D
16:59 kkeithley_ the things that pop into my head
17:01 kkeithley_ ದಯವಿಟ್ಟು
17:02 * kkeithley_ wonders how that looks in most IRC apps
17:03 hagarth kkeithley++: looks awesome in pidgin
17:03 glusterbot hagarth: kkeithley's karma is now 35
17:04 kkeithley_ google translate FTW
17:05 ndevos its horrible in irssi+screen :-/
17:05 kkeithley_ Unicode FTW
17:06 ndevos yeah, I think it is the rhel6 version of screen that does not like utf-8 very much
17:06 ndevos or the rhel6 irssi version for that matter
17:06 kkeithley_ smoke test failed posix tests? huh? Good old spurious failures
17:06 kkeithley_ Fedora FTW?
17:07 ndevos I run my irc client on a *server*
17:07 ndevos and ssh into that server from a Fedora laptop
17:07 ndevos the symbol table looks much nicer now :)
17:08 kkeithley_ it's going to pass this time
17:09 kkeithley_ yep
17:09 ndevos and I like the pub_ priv_ prefixes much better too
17:10 kkeithley_ where pub can mean public (visible) or published (declared in the header) ;-)
17:11 ndevos and priv_?
17:11 kkeithley_ private
17:11 ndevos oh, and a follow-up patch should definitely use macros for the .asm aliases
17:14 JustinClift kkeithley_: With the GlusterFS hardware we need putting in the DMZ, how much gear is there?
17:14 kkeithley_ agreed, mostly. a one-liner macro to replace a one-liner statement doesn't seem like a big win. The multi-line extern decl+attribute+asm symver clearly would be a win.
17:15 kkeithley_ one 2U server, two 3U raid enclosures. 8U total
17:15 JustinClift kkeithley_: Asking because we've just gotten a bunch of stuff for Ceph purchased and is currently being racked for their public usage.
17:15 JustinClift kkeithley_: k, that sounds like it's not a lot of gear
17:16 ndevos well, you have to type the whole thing often, GFAPI_PUBLIC("glfs_open", "3.4.0"); would be easier to do
17:16 kkeithley_ my thoughts exactly. But getting a response from IT is like pulling teeth. Actually, having just had a tooth pulled, getting a response from IT is a lot worse
17:17 JustinClift kkeithley_: Understood
17:17 JustinClift kkeithley_: My mgr is asking me what we need, so it can be made to happen (including hardware wise).
17:18 kkeithley_ ndevos: right, I'm not questioning that it's the right thing to do...
17:18 JustinClift Actually, I have an idea.  I should put you two together to discuss this, so I'm not johnny-in-the-middle.
17:18 * JustinClift gets that done instead
17:19 kkeithley_ JustinClift: sure, if he can make it happen. me. you. whoever.
17:20 kkeithley_ stepping out for lunch and to vote. biab
17:29 semiosis kkeithley_: http://imgur.com/ZoXylod <-- in Konversation
17:29 glusterbot semiosis: <'s karma is now -2
17:29 semiosis glusterbot: and not low enough!
17:32 ndevos kkeithley_: http://imgur.com/Go8sn3k <-- irssi+screen
17:32 glusterbot ndevos: <'s karma is now -3
17:34 JustinClift ndevos: For me, it was just question mark characters.  Not even slightly looking like it attempted unicode <stuff>.
17:34 hagarth Konversation++ :)
17:34 glusterbot hagarth: Konversation's karma is now 1
17:36 lalatenduM joined #gluster-dev
17:59 hchiramm joined #gluster-dev
18:26 kkeithley_ JustinClift: we have the hardware here in the lab in Westford. We just want to move it to the DC DMZ. Nothing to buy.
18:29 kkeithley_ misc said there's room in the RDU2 DC. Could never get an answer about space in the PHX2 DC. I just want IT's okay to ship it somewhere, with the knowledge that it'll get installed and not sit in the box(es) growing mold (or mould)
18:34 ndevos kkeithley_: it would be cool if they could be put in AMS ;)
18:36 kkeithley_ I'd happily put it there.
18:38 ndevos kkeithley_: something is broken in that libgfapi :-/
18:38 kkeithley_ hmmm
18:38 ndevos kkeithley_: I have a RHEL-6.6 installed and can run this with the rhel packages:
18:38 ndevos qemu-img create -f raw gluster://172.31.122.133/g​fapi/symbol-versioning.bin 2M
18:39 ndevos when I update the packages, I get an error
18:39 ndevos [2014-11-04 18:37:37.154115] E [glfs.c:191:create_master] 0-glfs: master xlator for gfapi initialization failed
18:40 ndevos that is the only error... not sure how to enable more verbose logging, and if that helps at all
18:40 * ndevos install ltrace
18:44 kkeithley_ xlator_set_type -> gf_strdup, and xlator_dynload, dlopen, several dlsyms.  bump logging to LOG_TRACE?
18:45 kkeithley_ I know, how to do that!
18:48 kkeithley_ I bet it's that in the "master" xlator (glfs-master.c) the requisite functions aren't public/exported from the shlib
18:50 kkeithley_ init, fini, options, fops, cbks, class_methods, etc.
18:50 kkeithley_ hmmm
18:51 ndevos I dont know, glfs_init() seems to return NULL:
18:51 ndevos 1949 glfs_init(0x7f514126a4f0, 0, 0x7f514123e908, 0x7f513f2d12a0, 0x7f514126b9c0)                       = 0xffffffff
18:51 ndevos or, rather -1
18:51 kkeithley_ yup, I'm not surprised.
18:51 ndevos but I do not see any details with ltrace
18:55 JustinClift kkeithley_: k, I'll hassle for that result then, and reject the "lets buy new hardware" argument
18:56 ndevos strace shows an mmap() and mprotect() on /usr/lib64/glusterfs/3.7dev/xlat... - no errors there either
18:57 ndevos I'm pretty sure there is a magic environment to debug libdl somehow
18:57 kkeithley_ I don't know ltrace. strace would only show the dlopen and dlsym of libgfapi.so.0, not the results of searches for symbols
18:58 ndevos ltrace for libraries, you can see what glfs_* functions are called
18:58 kkeithley_ the gfapi "master" xlator is in api.so
18:58 kkeithley_ ah, okay
18:59 ndevos glfs_init() is already returning the error
18:59 ndevos "man ld.so" has some interesting environment variables
19:03 kkeithley_ got it, I think
19:04 kkeithley_ master calls glfs_init_done, which is not exported. Need another private symbol exported
19:04 kkeithley_ not exported from libgfapi.so
19:05 kkeithley_ sigh
19:07 ndevos https://devos.fedorapeople.org/​tmp/qemu-img.dl-debug.log.2021 contains the debug
19:12 kkeithley_ nothing about the master xlator in api.so referencing glfs_init_done
19:19 kkeithley_ want to try patch set 12?
19:22 ndevos sure
19:22 ndevos whats the difference?
19:23 kkeithley_ glfs_init_done() is exported @@GFAPI_PRIVATE
19:24 kkeithley_ just saw Anders' comments about some typos in a couple files. I'll post patch set 13 in a minute
19:24 ndevos ah, yes, indeed api.so has a reference to that
19:26 kkeithley_ yup
19:35 ryao__ joined #gluster-dev
19:36 ndevos kkeithley_: how about a macro, like this? GFAPI_PUBLIC("glfs_init", "3.4.0")
19:36 ndevos with some sed, you can replace it: sed -r 's/^asm.*pub_(.*),.*(3\..\..).*/GFAPI_PUBLIC("\1", "\2");/'
19:37 ndevos one for pub_, one for priv_
19:39 kkeithley_ Next time I need a regular expression I know who to come to
19:45 ndevos kkeithley_: http://paste.fedoraproject​.org/147869/14151303/raw/ is patch with that sed applied, and the two macros added
19:46 ndevos it compiles, and the symbols look good
19:47 kkeithley_ yeah, I already had the macros in patch set 12 and 13. well, revised now. I just used your sed magic in my tree
19:49 * ndevos is in his living room, and running mock over wifi is *slooooooow*
19:50 kkeithley_ compiled, symbols look good. the acid test though. what does your qemu-image program do?
19:51 ndevos its almost done with rebuilding patch #13...
19:51 kkeithley_ gotta run and pick up my wife, biab
19:54 ndevos uh, no. patch #13 misses something too - not working :-/
19:54 * ndevos is afk for a bit too
19:55 ndevos maybe something went wrong, api.so still uses glfs_init_done?
19:56 JustinClift Hmmm
19:56 * JustinClift just noticed the GlusterFS Consultants page is online
19:56 JustinClift Anyone know if we did some kind of announce about that?
19:56 JustinClift If not, we need to
19:56 JustinClift xavih: ^
19:57 JustinClift xavih: Btw, DataLab should expand its entry out, to mention you're code contributors and such
19:58 * JustinClift could have sworn it wasn't supposed to be your contact details on there for the DataLab entry too, but <shrug>.
20:03 lalatenduM JustinClift, I have not see any announcement for it
20:28 ndevos kkeithley_: hmm, my build is missing the versioning for glfs_init_done... building again
20:28 kkeithley_ yes, priv_glfs_init_done is exported as glfs_init_done@@GFAPI_PRIVATE
20:29 ndevos yeah, I see that in my sed'ed version
20:29 ndevos but not in the package I've just installed - maybe I deleted a wrong one
20:30 ndevos building in mock over wifi takes 17(!) minutes
20:32 kkeithley_ why, is it chatty?
20:33 ndevos aha! now I see, I'm building for fedora-20-i386
20:34 ndevos it does not pickup the changes if you copy keep on copying the epel-6-x86_64 packages to the vm
20:34 ndevos I think it downloaded all the packages for the buildroot, or something like that
20:35 ndevos doing that over wifi, and saving them on a network mount does not make it faster
20:36 ndevos 100Mbit internet connection, that definitely makes my wifi the bottleneck - wired LAN in my office, much faster
20:37 kkeithley_ oh, indeed
20:38 kkeithley_ aren't the buildroot packages cached?
20:39 ndevos yes, I thought so... but the "yum update" step in mock seems to take quite long
20:45 ndevos kkeithley_: will you post a new version that uses the GFAPI_SYMVER_*_DEFAULT macros?
20:48 kkeithley_ yes, momentarily I will
20:49 kkeithley_ just checking that it builds first, one minute
20:50 ndevos ah, now I know whats up with mock, I'm hitting slow mirrors
20:50 ndevos or maybe thats my IPv6 over wifi that fails to roam
20:50 * ndevos looks for a wire
21:01 ndevos right, that makes a difference: INFO: Done(glusterfs-3.7dev-0.31​9.git0a1884d.fc20.src.rpm) Config(epel-6-x86_64) 3 minutes 20 seconds
21:03 kkeithley_ what did your GFAPI_PUBLIC macro look like? I seem to have forgotten how to write macros in C. brainfart!
21:03 ndevos +#define GFAPI_PUBLIC(symbol, version) asm(".symver pub_" symbol "," symbol "@@GFAPI_" version)
21:04 ndevos +#define GFAPI_PRIVATE(symbol) asm(".symver pub_" symbol "," symbol "@@GFAPI_PRIVATE")
21:04 ndevos but it passes "glfs_init" as stringm and "3.4.0" as string too
21:04 ndevos I think your macro does not require that, but I also did not try it
21:10 ndevos kkeithley_: patch #13 + my sed+defines fails - priv_glfs_subvol_done in api.so does not carry a version?
21:10 * ndevos builds a clean #13 now
21:11 kkeithley_ yup, Anders found that too.
21:11 ndevos oh, okay, so I dont need to test a clean #13 then
21:16 kkeithley_ maybe 14 should be good, less the symver macros
21:16 * kkeithley_ hopes
21:16 kkeithley_ I hate doing things in a rush
21:21 ndevos JustinClift: nice, all consultants are un Europe :D
21:31 ndevos kkeithley_: #14 IS ALIVE!
21:31 kkeithley_ it works?
21:33 ndevos yes
21:36 kkeithley_ whew, what an exercise
21:39 ndevos yeah, quite
21:40 ndevos kkeithley_: will #15 have the macros too?
21:42 kkeithley_ yup
21:42 ndevos nice!
21:49 kkeithley_ I hope that's a promise I can keep
21:51 kkeithley_ ;-)
22:03 kkeithley_ and for my next trick, change set 15 with macros
22:10 ndevos rebuilding again
22:16 ndevos kkeithley_: #15 works for me too
22:19 badone_ joined #gluster-dev
22:21 ndevos kkeithley_: does gfapi.map.in really need to be a .in file? I do not think there is any substition in there?
22:39 ndevos oh, an other issue, the nfs server uses glfs_resolve_at:
22:39 ndevos [2014-11-04 22:37:42.755496] W [xlator.c:191:xlator_dynload] 0-xlator: /usr/lib64/glusterfs/3.7dev/xlator/nfs/server.so: undefined symbol: glfs_resolve_at
22:40 ndevos I think the snapshot server uses that too :-/
22:45 kkeithley_ there's no substitution (yet). Maybe there never will be.  I tried adding EXTRA_DIST = gfapi.map to api/src/Makefile.am, and it worked in my local tree, but didn't work in the jenkins build
22:46 kkeithley_ worked meaning it was added to the source tarball from `make dist`
22:47 kkeithley_ why on earth is nfs/server using anything from gfapi?
22:48 ndevos it's doing a lookup without a transport connected, some workaround for handling UDP
22:49 ndevos api.so also has a reference to glfs_graph_setup
22:49 ndevos the snapshot-*.so do not seem to have references to unversions symbols
22:54 kkeithley_ glfs-master.c has its own definition of glfs_graph_setup()
22:54 kkeithley_ not sure why it has glfs_ prefix. Should I fix it while I'm here?
22:55 ndevos sure, feel free to
22:56 kkeithley_ there is no other definition (or declaration) of glfs_graph_setup()
22:56 ndevos ah, no, thats my mistake
22:57 ndevos glfs_graph_setup() just is defined in api.so, my quick readelf loop that tries to detect unversioned glfs_* functions cought it incorrectlt
22:58 kkeithley_ yup, no problem, better to check
22:59 kkeithley_ and good that you checked and found glfs_resolve_at
23:01 ndevos the nfs-server uses others too, they are in the gerrit comment
23:01 kkeithley_ oh, just saw your comment on patch set 15. We need glfs_free_fro....
23:01 kkeithley_ yeah
23:03 ndevos okay, and those are really the only ones in the xlators we package (including glusterfs-extra-xlators)
23:07 kkeithley_ seems like using glfs_anything in xlators and heal is just drilling a hole to China.
23:07 kkeithley_ Let's just punch holes in our abstraction layers where ever it suits us
23:08 kkeithley_ ugly
23:09 ndevos yes, I agree, and I do not really like it either
23:10 ndevos unfortunately libgfapi has some advanced functions (like resolve_at) that are not available in the private libs
23:11 kkeithley_ although if we weren't switching to ganesha, I'd say the NFS server ought to be rewritten with gfapi.
23:11 ndevos I only accepted its usage in the nfs-server, because other xlators use libgfapi too...
23:11 ndevos ah, I still think the fuse client should use libgfapi
23:12 ndevos I do not plan to rewrite the nfs-server, but the fuse client is an option
23:16 kkeithley_ yup
23:17 kkeithley_ then it should be moved to libglusterfs, like the comment in nfs/server/src/mount3.c says about glfs_new_from_ctx(),
23:19 kkeithley_ s/glfs_new_from_ctx/glfs_resolve_at/
23:20 systemonkey joined #gluster-dev
23:21 ndevos yes, that was my suggestion when I 1st NACKed that change
23:21 kkeithley_ patch set 16
23:21 kkeithley_ oho, did you really
23:22 kkeithley_ good on you
23:22 ndevos but, making that change was not something trivial :-/
23:22 kkeithley_ and it went through anyway. :-(
23:22 ndevos and, he had a point that other xlators also used libgfapi already...
23:22 kkeithley_ well, there's never time to do it right the first time
23:23 kkeithley_ well, we're not going to fix it here/now.
23:24 ndevos not me, not now, no
23:24 ndevos but I'll run a last test with #16 now
23:25 kkeithley_ If the other xlators are using _public_ APIs, I care less, perhaps, about that.
23:26 kkeithley_ Since we're putting use of these non-public APIs under the spotlight, maybe something good will come from that
23:27 ndevos well, glfs_resolve_at could be public, I think? glfs_new_from_ctx and glfs_free_from_ctx probably should not
23:27 kkeithley_ A good Google Summer of Code project might be to use a link map (with or without versions) on some of the other libraries
23:28 ndevos yeah, that would be nice
23:28 kkeithley_ with or without symbol versions
23:31 ndevos yay, with #16 nfs works too
23:31 ndevos and qemu-img did not break
23:36 kkeithley_ excellent
23:36 ndevos kkeithley_: I'll drop off now, we'll catch up in tomorrows meeting
23:37 kkeithley_ cheers, I owe you big time. Anders too. Should figure out how to reward hime
23:37 kkeithley_ him
23:37 kkeithley_ thanks for all the help
23:37 ndevos yes, lets think about getting him something
23:37 ndevos you're welcome, and thank you too!
23:37 ndevos kkeithley++
23:37 glusterbot ndevos: kkeithley's karma is now 36

| Channels | #gluster-dev index | Today | | Search | Google Search | Plain-Text | summary