2017-02-02 GnuCash IRC logs
00:15:27 *** O01eg has quit IRC
00:35:54 *** CharlesIC has joined #gnucash
00:35:57 <CharlesIC> hi please help
00:36:04 <CharlesIC> im going crazy trying to reconcile something
00:52:37 *** quakestring has quit IRC
01:07:58 *** CharlesIC has quit IRC
01:19:56 *** fell has joined #gnucash
01:22:17 *** fell__ has quit IRC
01:26:58 <movedx> warlord: This website possibly isn't considered suitable for high availability, but if it is, then I would expect to see two IPs for the purposes of hitting one of two web servers directly or hitting one of two load balancers.
01:27:28 <movedx> warlord: If the website is static, I would consider throwing it up on GitHub Pages: free for static sites.
02:30:14 *** gjanssens has joined #gnucash
02:30:15 *** ChanServ sets mode: +o gjanssens
02:34:54 *** cartsoftware has quit IRC
02:35:28 *** cartsoftware has joined #gnucash
02:38:04 *** cartsoftware1 has joined #gnucash
02:38:31 *** cartsoftware has quit IRC
02:38:32 *** cartsoftware1 is now known as cartsoftware
03:06:54 *** cira has quit IRC
03:07:14 <gjanssens> .
03:07:14 <gncbot> gjanssens: Sent 6 hours and 42 minutes ago: <warlord> Response was Wietse is that he wrote that program 20 years ago and not to expect any change.
03:07:15 <gncbot> gjanssens: Sent 6 hours and 36 minutes ago: <warlord> Regarding adding jralls, does he have write access to the admin repo? If so, then sure.
03:10:22 <gjanssens> warlord: good we created our own solution then for postfix
03:11:01 <gjanssens> warlord: jralls does have write access, so I have added him to the list of people to receive notification mails
03:15:48 <gjanssens> I've just added another change to send only one mail instead of 2. For the admin repo we don't need with and without diffs.
03:16:14 <gjanssens> This will only be in effect starting from the next push to the repo
03:23:16 *** mrklintscher has joined #gnucash
03:26:02 *** cartsoftware has quit IRC
03:35:46 *** fabior has joined #gnucash
03:49:33 *** fabior has quit IRC
03:57:10 *** cartsoftware has joined #gnucash
04:04:26 *** fabior has joined #gnucash
04:05:47 *** cyphase_ has joined #gnucash
04:06:12 *** cyphase has quit IRC
04:06:15 *** cyphase__ has joined #gnucash
04:42:22 *** cartsoftware has quit IRC
05:03:49 *** cira has joined #gnucash
05:21:38 *** cartsoftware has joined #gnucash
06:59:45 *** fabior has quit IRC
07:00:43 *** Jimraehl1 has joined #gnucash
07:01:29 *** Jimraehl1 has quit IRC
08:40:15 *** fabior has joined #gnucash
08:43:27 *** fabior has quit IRC
08:57:49 *** fabior has joined #gnucash
09:09:42 <warlord> movedx: it's not static.
09:10:08 <warlord> gjanssens: sounds good.
09:14:56 *** puck has quit IRC
09:19:57 *** puck has joined #gnucash
09:20:55 *** kael has joined #gnucash
09:47:30 *** fabior has quit IRC
09:52:25 *** kael has quit IRC
10:01:10 *** fekepp has quit IRC
10:01:50 *** fekepp has joined #gnucash
10:05:07 *** fabior has joined #gnucash
10:24:58 *** quakestring has joined #gnucash
11:11:17 *** O01eg has joined #gnucash
11:14:11 *** quakestring has quit IRC
11:19:03 *** kael has joined #gnucash
11:29:29 *** fabior has quit IRC
11:30:13 *** storyjesse has quit IRC
11:44:33 *** fabior has joined #gnucash
12:00:10 *** quakestring has joined #gnucash
12:06:49 *** quakestring has quit IRC
12:26:33 *** Mechtilde has joined #gnucash
12:44:53 *** Kreisel has joined #gnucash
12:45:25 *** Kreisel has left #gnucash
12:45:59 *** kriesel has joined #gnucash
12:59:32 *** kriesel has quit IRC
13:18:52 *** mrklintscher has quit IRC
13:29:44 *** fabior has quit IRC
13:39:07 *** mlncn has quit IRC
13:39:39 *** KaiForce has joined #gnucash
13:51:30 <gjanssens> jralls: thanks for your feedback on the gtk3 port
13:52:25 <jralls> gjanssens: That's a lot to bite off this close to the end of the cycle, but none of the alternatives are attractive at all.
13:52:48 <gjanssens> Indeed
13:53:23 <jralls> Maybe longer term the right thing to do is to just draw the stupid reports with no HTML.
13:53:48 <jralls> But that's probably more work than converting to Gtk3.
13:53:59 <gjanssens> :)
13:54:10 <gjanssens> Yeah, not something for short term
13:54:55 <gjanssens> Anyway, I think the goocanvas route is likely the least intrusive
13:55:34 <jralls> So let's set priorities. I've almost to the point of creating a PR for numerics. It's not really done yet, but it's close enough to be usable.
13:55:37 <gjanssens> Though I'm tempted to continue on the gtktreeview route after all
13:56:21 <gjanssens> Well I'm currently ready to finish the last bits on my csv importer work
13:56:34 <gjanssens> The gtest stuff I pushed today was part of that
13:56:51 <gjanssens> After that it's Gtk3 for me
13:56:57 <jralls> Again, long-term I think that GtkTreeView is the better choice, but I doubt we can get it fixed in the 10 months remaining.
13:57:35 *** mlncn has joined #gnucash
13:57:45 <gjanssens> I'll make my final decision when I dig in.
13:57:56 <jralls> I still need to do some cleanup in the SQL backend. I think I'll not take on converting the XML backend to pure C++. I don't think there would be a lot of gain there.
13:58:19 <gjanssens> In the past I had an attempt to switch to goocanvas, but it wasn't very successful
13:58:39 <gjanssens> It's been a while so I don't remember very well what my problems were
13:58:50 <gjanssens> Ok
13:58:55 <jralls> I hope I can get that done in a month or two.
13:59:54 <jralls> If goocanvas won't work then we may have to just draw directly on a Cairo canvas. It's a bit more work, but we can probably find code to steal in Gtk itself.
14:00:17 <gjanssens> Right
14:00:49 <gjanssens> I think goocanvas can work. I think I was mostly hindered by the confusing organisation of our register code.
14:01:17 <jralls> Anyway, after cleaning up the SQL backend I'll defer further work on C++ conversion to next year and dive in with you on Gtk3/Webkit2/canvas.
14:01:26 <gjanssens> It was a quick & dirty first stab back then. I'll pay more attention this time :)
14:01:31 <gjanssens> Good
14:02:42 <gjanssens> I don't think I'll need a month or two for the csv importer, so with some luck I can have some gruntwork done by the time you finish the sql backend cleanups
14:03:18 <jralls> OK.
14:04:24 <gjanssens> Also related to the 2.7/2.8 cycle, but not gtk related: what's the status of the cmake build system codesmythe has been working on ?
14:05:05 <gjanssens> We've been using it on Windows for a while but what is missing to make this our primary build system for 2.8 ?
14:05:21 <gjanssens> (If we want to and have time to finish it of course)
14:06:16 <jralls> I've been using it as my primary build for developing. It's tons faster and it's much easier to run individual test programs because it doesn't have all of the bleeping extra paths that have to be configured.
14:07:14 <gjanssens> I had to reconfigure this morning because I switched branches, and decided to give it a go as well
14:07:25 <jralls> All that's needed to ditch autotools is make distcheck. Codesmythe said a couple of months ago that he'd set that up, I haven't heard anything from him since.
14:07:45 <gjanssens> It's indeed much faster (I'm using ninja-build like in the windows build)
14:08:02 <gjanssens> Ok, hope he can finish this.
14:08:32 <gjanssens> But I'm having some mild issues.
14:08:48 <gjanssens> On a clean build I'm getting lots of
14:08:50 <gjanssens> gnc.module-WARNING **: Failed to dlopen() '/kobaltnet/janssege-development/builds/gnucash/master/lib/gnucash/libgncmod-csv-import.so': /kobaltnet/janssege-development/builds/gnucash/master/lib/gnucash/libgncmod-csv-import.so: undefined symbol: gnc_csv_col_type_strs
14:09:16 <gjanssens> I think they come from guile2 compile
14:09:29 <gjanssens> Which interprets the scm files
14:09:40 <gjanssens> These in turn do a lot of dlopening
14:10:06 <gjanssens> It looks like this is happening while libgncmod-csv-import.s isn't fully written yet
14:10:23 <gjanssens> And it's not always libgncmod-csv-import.so
14:10:34 <gjanssens> This is what I see on my cpp branch
14:10:52 <gjanssens> Before on master it was another library, also libgncmod
14:11:34 <gjanssens> It looks like this is not fatal but I don't like the warnings either
14:11:51 <gjanssens> Did you ever get similar warnings on a clean build ?
14:12:24 <jralls> Not that I've noticed, let me try one. Won't take long.
14:14:00 *** kael has quit IRC
14:14:27 * gjanssens is looking back at the earlier build log today
14:14:40 <gjanssens> In that build it happened with libgncmod-business-gnome.so
14:18:11 *** kriesel has joined #gnucash
14:18:29 *** kriesel has left #gnucash
14:22:13 <jralls> I don't see anything like that. I see deprecation warnings, for 'register' in swig-runtime.h, for scm_internal_stack_cache in gfec.c, for g_settings_list_keys in gsettings.h.
14:22:38 <gjanssens> Yes, I see those as well.
14:23:57 <jralls> Some errors compiling stylesheets.go about gnc_get_default_report_font_family, assertion 'top_list' != NULL (that will be a g_return_if_fail), and "Could not locate module gnucash/tax/de_DE interface v.0" for de_DE.go.
14:26:48 <jralls> Maybe ltdlopen (which guile uses a lot) is finding your "regular" gnucash libraries?
14:27:10 <gjanssens> No, because I don't have a regular gnucash installed any more
14:27:17 <gjanssens> Exactly to avoid such issues
14:29:41 <jralls> Hmm. /me fires up his fedora VM...
14:30:15 <gjanssens> Hmm, looks like I have some debugging to do... When building with ninja, my new csv import module isn't loading in gnucash for the same reason
14:30:54 <gjanssens> jralls: no need to fire your Fedora VM just yet :)
14:31:00 <gjanssens> eh fire up
14:31:14 <jralls> Too late, it doesn't take long! ;-)
14:31:36 <jralls> But are you seeing the problem on master builds or just on your cpp branch?
14:32:49 <gjanssens> I saw a similar link warning on master, the warning there was about libgncmod-business-gnome.so
14:33:20 <gjanssens> The libgncmod-csv-import.so warning is on my cpp branch and the undefined symbol error looks genuine
14:33:40 * gjanssens goes back to check the full warning on master...
14:34:30 *** codesmy82 has joined #gnucash
14:35:22 <gjanssens> jralls: The warnings on master are slightly different:
14:35:25 <gjanssens> (process:1582): gnc.module-WARNING **: Failed to dlopen() '/kobaltnet/janssege-development/builds/gnucash/master/lib/gnucash/libgncmod-business-gnome.so': /kobaltnet/janssege-development/builds/gnucash/master/lib/gnucash/libgncmod-business-gnome.so: file too short
14:36:28 <gjanssens> Had another file too short in an even earlier build today for libgncmod-customer_import.so
14:36:43 <gjanssens> These seem artifacts from parallel building.
14:37:13 * gjanssens wonders if codesmy82 is codesmythe
14:38:57 <codesmy82> gjanssens: yep, that's me. Came in to say that my plan is to have the make distcheck stuff done in the next couple of weeks.
14:39:45 <gjanssens> codesmy82: wonderful \o/
14:39:47 <codesmy82> Also, the file too short is parallel race conditions as you say.
14:40:04 <gjanssens> Is there a way to avoid it ?
14:40:56 <codesmy82> I did look into it when I did the original work. They don't affect anything, but can't quite remember why I concluded they were harmless.
14:42:06 <gjanssens> Ok
14:42:27 <codesmy82> I think the guile compile is trying to open a dependent .so that is not yet built. But since they are built by runtime things are okay.
14:43:18 <codesmy82> Seems to show up on Linux and not Mac or Windows.
14:43:22 <jralls> That sounds like there are missing dependencies.
14:43:32 <gjanssens> As long as the missing .so's are not required to resolve all symbols for scm I suppose that's true
14:43:47 <gjanssens> I'm more concerned about my other warning and the resulting omission of the csv impoter from gnucash
14:44:47 <gjanssens> The missing symbol gnc_csv_col_type_strs is declared extern in one header file and then defined in a related source file
14:45:05 <jralls> Ah, I do get the "libgncmode-business-gnome.so file too short" error on th eF25 VM.
14:45:45 <gjanssens> This has always worked fine using autotools, but now with cmake and ninja is seems to fail
14:47:28 <codesmy82> I'll refresh my memory about why the file too short thing is not an issue and report back later this week.
14:48:22 <codesmy82> gjanssens: I'm not at my computer so I can't help debug warnings/errors at the moment, but can later.
14:49:28 <gjanssens> codesmy82: that's ok
14:50:05 <gjanssens> If I find out myself, I'll report back. Otherwise I'll be glad to accept your help
14:56:41 * gjanssens is rebuilding using autotools for comparison... takes a little longer (yawn)
14:58:01 <jralls> It's on txf-de_DE.go. I seem to have moved it to equile-gnc.go by adding gncmod-ledger-core gncmod-business-gnome to GUILE_DEPENDS in src/report/locale-specific/us/CMakeLists.txt
15:02:07 <gjanssens> jralls: as I'm relatively new to cmake and ninja, which files do you check to find out which build step emits these warnings ?
15:04:12 <jralls> I'm trusting that ninja keeps its output together, so if it says Generating .../foo.go and then emits the error that the error comes from foo.go.
15:04:53 *** kael has joined #gnucash
15:06:46 <jralls> If that's not the case one would have to say 'ninja-build -j 1' to force it to do one thing at a time.
15:08:07 <gjanssens> jralls: ok
15:08:21 <gjanssens> I am not really sure if that's the case yet
15:08:46 <gjanssens> In my windows builds this morning I had errors with compiling gtest sources
15:08:59 <gjanssens> They appeared to come from an earlier build step
15:09:17 <gjanssens> I say "appeared" because didn't investigate too deeply.
15:09:34 <gjanssens> I'll assume for now ninja does as you say :)
15:10:15 <jralls> Well, I'll test it. I launched the next try with -j 1.
15:10:42 <jralls> It's a bit of whack-a-mole chasing down the missing dependencies.
15:15:30 <jralls> Well, that run didn't complain at all about incomplete files.
15:15:42 <gjanssens> Yeah, there are more enjoyable things to do...
15:16:13 <gjanssens> Heh, probably not, you're no longer building in parallel
15:16:27 <jralls> Like debugging gnc-numeric test failures...
15:16:30 <gjanssens> So there won't be any incomplete libraries
15:16:46 <gjanssens> Either the library is built or it's not there
15:17:02 <gjanssens> the dlopen call comes from gnc_module IMO
15:17:24 <gjanssens> which searches for all libaries whose name starts with libgncmod
15:18:26 <gjanssens> So it's not even sure these libraries really are dependencies
15:18:37 <jralls> Ah, so it's just finding filenames that start with libgncmod that are still under construction but aren't actually needed?
15:19:32 <jralls> That's plausible. So one work-around would be to change the build order so that all of the modules are built before compiling any scheme.
15:19:47 <codesmy82> Ah, that sounds familiar.
15:20:16 <jralls> OTOH, if it's just noise and not a real error maybe it makes sense to ignore the problem.
15:20:39 *** fabior has joined #gnucash
15:21:43 <jralls> Another would be to get rid of gnc-module altogether. AFAICT it doesn't actually do anything for us except slow down loading. dlopening is tons more expensive than dynamic linking.
15:23:56 <gjanssens> Completely agree.
15:24:10 <gjanssens> Shall we do that first? ;)
15:24:50 <gjanssens> So the autotools build finised and it does indeed work ok
15:25:24 <gjanssens> So I start looking for build configuration differences...
15:39:19 <jralls> Might be a ninja vs. make difference rather than cmake vs. autotools. Make isn't anywhere near as parallel.
15:41:09 <jralls> Just tried cmake -G make && make and don't see any of the incomplete library errors.
15:43:35 <jralls> But if I say `make -j 20` I do.
15:45:27 <jralls> Now trying autotools with make -j 20
15:46:49 <gjanssens> Well duh, my issue with the missing symbol is of the too obvious variety...
15:46:58 <gjanssens> I missed a source file in CMakelists.txt
15:47:11 <gjanssens> Which I had properly added to Makefile.am
15:47:37 <gjanssens> Leaves only the file too short ones
15:47:46 <jralls> Ah, I thought you were still on the partly-built library problem.
15:49:18 <gjanssens> Sorry I have been jumping up and down between the two threads...
15:49:36 <jralls> make -j 20 with autotools fails trying to link libgnc-qof.so.
15:49:49 <gjanssens> codesmy82: my issue with the missing symbol turned out to be pebkac
15:50:00 *** Mechtilde has quit IRC
15:50:16 <gjanssens> jralls: let me try that here as well
15:55:27 <gjanssens> make -j20 completes without errors here
15:55:34 <gjanssens> And no file too short warnings
15:56:23 <gjanssens> I think we should just let this rest for now
15:56:47 <gjanssens> I am tempted to conclude 'file too short' is harmless
15:57:22 <jralls> I see that now too. I'd tried to pass a relaative path for gtest/gmock, which of course worked for configure and cmake but not for autotools which cds to every subdir.
15:57:46 <gjanssens> Oh right
16:02:11 <warlord> Maybe once I get my network upgraded I can have code do double-duty as www as well (as a mirror from Linas).
16:07:30 <jralls> It looks to me like the recursive build prevents autotools from having the file too short problem as that invokes make on one dir at a time.
16:07:54 <gjanssens> That would make sense indeed.
16:08:05 <gjanssens> Anyway, I have to leave for tonight.
16:08:11 <gjanssens> See you tomorrow!
16:08:28 <jralls> warlord: Having a fallback would be good. Maybe linas could provide a fallback for the wiki and mailing lists as well.
16:08:36 <jralls> G'night, Geert.
16:09:36 <jralls> I'll have the numeric PR ready for you when you get up!
16:09:49 <gjanssens> Yay
16:09:51 *** kael has quit IRC
16:11:42 *** kael has joined #gnucash
16:12:15 *** gjanssens has quit IRC
16:22:22 *** codesmy99 has joined #gnucash
16:23:27 *** codesmy82 has quit IRC
16:35:06 *** kael has quit IRC
16:51:14 *** O01eg has quit IRC
16:51:46 *** kael has joined #gnucash
17:27:27 *** mlncn has quit IRC
17:35:43 *** kael has quit IRC
17:38:08 *** kael has joined #gnucash
17:42:20 *** fabior has quit IRC
17:44:23 *** kael has quit IRC
17:57:53 *** fabior has joined #gnucash
18:02:14 *** KaiForce has quit IRC
18:03:03 *** mrklintscher has joined #gnucash
18:21:16 *** kael has joined #gnucash
18:48:57 *** fabior has quit IRC
18:54:26 *** mrklintscher has quit IRC
18:55:39 *** kael has quit IRC
19:17:36 *** yuriks has quit IRC
19:25:43 *** yuriks has joined #gnucash
19:39:51 <warlord> jralls: at one point we did have a backup MX record, but there's no good way to have the lists on multiple machines. The website is quazi-static without any backend services so it's easy to replicate.
19:40:34 <warlord> But for mailman I think you need to share a backend storage ... I don't know how well GlusterFS would work for something like that.
19:51:10 *** mlncn has joined #gnucash
19:55:36 *** codesmy80 has joined #gnucash
19:55:38 *** codesmy99 has quit IRC
19:56:08 *** codesmy80 has quit IRC
19:58:43 *** mlncn has quit IRC
21:18:04 *** kael has joined #gnucash
22:38:34 *** O01eg has joined #gnucash
23:01:22 *** O01eg has quit IRC
23:14:45 *** kael has quit IRC
23:15:36 *** O01eg has joined #gnucash
23:59:07 *** Robert has joined #gnucash
23:59:24 *** Robert has quit IRC