16:02:29 #startmeeting fpc 16:02:29 Meeting started Thu Aug 6 16:02:29 2015 UTC. The chair is geppetto. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:02:29 Useful Commands: #action #agreed #halp #info #idea #link #topic. 16:02:29 #meetingname fpc 16:02:29 The meeting name has been set to 'fpc' 16:02:29 #topic Roll Call 16:03:17 geppetto limburgher mbooth orionp racor Rathann SmootherFr0gZ tibbs tomspur: FPC ping 16:03:29 .hello mbooth 16:03:30 mbooth: mbooth 'Mat Booth' 16:03:37 #chair mbooth 16:03:37 Current chairs: geppetto mbooth 16:04:28 hi 16:04:29 #chair Rathann 16:04:29 Current chairs: Rathann geppetto mbooth 16:04:34 #chair tomspur 16:04:34 Current chairs: Rathann geppetto mbooth tomspur 16:05:23 switched to a balcony seat today ;-) 16:05:41 still at the back though ?:) 16:05:51 ROFL.... yup 16:06:03 try to keep a low profile 16:06:12 which mtg is this now? 16:06:29 Corey84: FPC 16:07:00 * Corey84 sits in back row to learn about as he is not in FPC presently 16:08:03 tibbs|w: was around a few minutes ago, so I think we could start 16:08:29 Yeh, I figure he'll get here 16:08:29 Yeah, I howdy'd earlier. 16:08:32 ok 16:08:35 #chair tibbs 16:08:35 Current chairs: Rathann geppetto mbooth tibbs tomspur 16:08:49 #topic Schedule 16:08:54 https://lists.fedoraproject.org/pipermail/packaging/2015-August/010905.html 16:09:05 #topic #555 Copylib bundling request: kwsys in castxml 16:09:05 .fpc 555 16:09:05 https://fedorahosted.org/fpc/ticket/555 16:09:06 geppetto: #555 (Copylib bundling request: kwsys in castxml) – fpc - https://fedorahosted.org/fpc/ticket/555 16:10:24 There's not much anyone can do here. 16:10:31 nice … so "bundling library" which is giant 16:10:56 Yeah. 16:11:17 It's a bunch of files, but the source is mostly ifdefs and such to do platform abstraction. 16:11:18 contains replacements for most of libc++? 16:11:39 I don't understand why the namespace is so important, though. 16:12:08 I assume so that N libraries can bundle it all within different namespaces 16:12:15 shared libraries that is 16:12:23 I mean, why do they not just stick it in its own namespace instead of having packages renamespace it? 16:12:34 And then stick that in a shared library.... 16:13:01 geppetto: which is still code duplication :( 16:13:01 because it's more fun if you have N different bundles which are all slightly different versions 16:13:49 huh, this is the same upstream as CMake 16:14:03 and VTK and ParaView... 16:14:28 and InsightToolkit 16:15:27 I assume they did this because they want to support many platforms 16:15:35 but still it should be a shared library 16:16:12 It seems while renamespacing it you can select which parts you want to include, so that each bundling is different from the final content 16:16:25 yeh, I'm pretty heavy -1 16:16:35 tomspur: yes, but does it make sense in Fedora? 16:17:28 Rathann: I'm wondering why that cannot be solved with normal ifdefs... 16:18:38 depending on the number of platform-specific features you depend on, this can get unwieldy 16:18:54 so I can understand why they wanted to write an abstraction layer 16:19:38 Yeh, but this is basically a c++ version of glib … it's insane to bundle it in every piece of code that wants to use it 16:19:47 yep 16:20:29 ha … the regexp. code isn't a wrapper … it's a full implementation 16:21:08 huh 16:21:17 NIH syndrome? 16:21:34 there might not be a system one on win32 16:23:03 I mean … I can see the desire to just include one, just need to face up to the fact you are a real library/API at that point 16:23:25 reimplements parts of STL too 16:25:11 and string implementation is broken on non-C locales (has a hardcoded table of uppercase letters) 16:25:23 meh 16:25:26 I'm seeing this as a really bad idea, but I'm not sure that there's any reasonable solution that doesn't involve just pulling the packages. 16:26:01 yeh, I mean someone could try to package it as a shared library in Fedora 16:26:19 Yes, but how would other software use it? 16:26:30 Choose a semi random namespace and just do it … but meh. 16:26:46 It's almost better to say "upstream insane, port to boost" 16:26:54 it looks like it's even actively maintained 16:26:57 I have no idea how that would work, or how C++ namespacing works at all. 16:27:01 last commit was 2 days ago 16:27:13 It wasn't even part of the language when I first learned it in school. 16:27:48 huh 16:27:55 Rathann: Are you sure you are looking at just commits to the kwsys bits and not castxml in total? 16:28:06 geppetto: I'm looking at http://public.kitware.com/gitweb?p=KWSys.git;a=log 16:28:13 this is kwsys upstream, isn't it? 16:28:25 ahh, well … that's much worse then as the castxml bits haven't been updated in months/years 16:28:40 they seem to support such exotic platforms as HP-UX and VMS 16:28:51 Sure, why not? 16:29:08 on the upside I'm sure nobody cares if you bundle everyone on HP-UX 16:29:09 But don't make people trying to do reasonable software engineering suffer because of it. 16:30:08 So, this is a lot like the scintilla problem. 16:30:32 It's bundled several times but there's no standalone package which someone could use. 16:30:59 And until we get a volunteer to make that package, we can't tell the bundlers to use it. 16:32:30 yeah 16:33:06 I wonder... does cmake bundle it? ;) 16:33:26 looks like debian castxml bundles it 16:33:26 It's sort of like the jquery issue. Once it was packaged we started making people use it. 16:34:27 So, what to do? Temporary exemption until we have a separate package? 16:36:28 I guess … although that doesn't seem very temporary 16:37:39 Well, the idea is that if we put a limit on it then someone will be motivated. 16:38:01 hm hold on, castxml is not in Fedora yet 16:38:26 so the submitter could be asked to package kwsys 16:38:40 yeh, but cmake is 16:38:47 and cmake bundles it, I assume 16:38:53 It's bundled in debian 16:39:46 * Rathann checks 16:40:03 Uh: http://www.cmake.org/gitweb?p=cmake.git;a=tree;f=Source/kwsys it does 16:40:14 it does 16:40:29 yeh, even more amusing 16:40:29 *sigh* 16:40:33 From the cmake specfile: 16:40:34 # Source/kwsys/MD5.c 16:40:34 # see https://fedoraproject.org/wiki/Packaging:No_Bundled_Libraries 16:40:35 Provides: bundled(md5-deutsch) 16:40:46 *LOL* 16:41:21 I mean … sure, just ignore all the other files in that dir. :) 16:41:41 missing the elephant in the room... 16:41:49 I have a suspicion that orion might be interested in maintaining a separate package. 16:42:16 Although I guess if cmake is the same upstream they might feel differently about it … *sigh* 16:42:25 still... existing bundling is no excuse for new bundling 16:42:29 tibbs: that would be cool 16:42:59 Rathann: Yeh, I'm just saying that while I don't know how useful castxml is … it's not like we can bounce cmake :) 16:45:35 so... -1 and recommend packaging kwsys separately (ask orionp for help)? 16:47:45 yeh, I guess … although it's somewhat complicated due to cmake 16:48:14 If something big/useful is stuck behind castxml, I could see trying the temp. exception thing 16:48:26 Agreed. -1 for now but I would agree to a "check back in two releases" exemption. 16:49:00 * tomspur nods 16:49:33 We do need to file bugs for tracking. 16:49:38 yeh 16:49:56 And it would be kind of nice if we knew that it would be possible to unbundle once there's a standalone package. 16:50:05 I mean, I don't know if that's remotely reasonable. 16:51:50 How much patching is involved in using the library in another namespace? How many of these bundlers have made modifications to the code base? 16:52:28 It's interesting that git and submodules have made it easier to bundle, but also easier to keep track of local modifications. 16:54:06 yeh, java infecting everyone has "helped" too :) 16:54:09 * geppetto shrugs 16:56:15 #action The opinion seems to be that this is a mess, -1 for bundling unless you can come up with an important usecase for castxml getting into Fedora soon. 16:56:44 #action orionp Do you want to help with trying to create a real library? 16:57:34 I'm filing a bug against cmake and others 16:57:41 cool, much thanks 16:58:31 #topic Open Floor 16:58:54 I don't think we need to say more than tibbs did on 556 16:59:01 So anyone have anything else? 16:59:20 One thing to note is that next weeks meeting is almost certainly canceled 17:01:00 https://bugzilla.redhat.com/show_bug.cgi?id=1251198 17:01:09 Ok, I'll close the meeting in a couple of minutes … see everyone in a couple of weeks 17:01:14 Rathann: thanks again 17:01:25 Wow, nothing else. 17:01:34 Well, we're down to very few tickets now. 17:01:35 No, we only had the one new ticket 17:01:39 * geppetto nods 17:01:55 yeah, congratulations on getting through the pile of tickets ;) 17:02:00 over the last few months 17:02:07 Thanks. 17:02:18 I could probably clean up a few more. 17:02:19 I have to leave now as well, so I'll file bugs against other bundlers later tonight 17:02:23 yeh, this feels like the first meeting that only took an hour that we've had this year :) 17:02:28 \o/ 17:02:32 I certainly don't want to extend it. 17:02:35 also, it's holiday season ;) 17:02:46 yeh, true 17:02:48 so take care, get some rest, etc. 17:02:52 * geppetto nods 17:02:55 bye 17:02:56 It's "don't go outside because it's 40C" season for me. 17:03:16 tibbs: vacation somewhere colder ?;) 17:03:28 Well, Rochester next week. 17:03:42 yeh, weather is looking sucky too :( 17:03:51 My wife just got back from Norway (where it didn't get above 10C) but I didn't go this year. 17:04:28 Your vacation was alone time ?;) 17:04:48 Pretty much. Which means I stayed in the office until midnight every day. 17:05:03 #endmeeting