2017-09-05 GnuCash IRC logs

00:53:10 *** Mechtilde has joined #gnucash
00:53:34 *** User has joined #gnucash
00:56:44 *** User has quit IRC
01:01:29 *** jotrago has quit IRC
01:24:35 *** O01eg has quit IRC
01:58:30 *** Mechtilde has quit IRC
02:19:58 *** pilotauto_ has joined #gnucash
02:20:08 *** pilotauto has quit IRC
02:28:15 *** jotrago has joined #gnucash
02:58:10 *** jotrago has quit IRC
03:00:09 *** jotrago has joined #gnucash
03:01:46 *** fabior has joined #gnucash
03:10:12 *** gjanssens has joined #gnucash
03:10:12 *** ChanServ sets mode: +o gjanssens
03:10:32 <gjanssens> .
03:10:32 <gncbot> gjanssens: Sent 7 hours and 42 minutes ago: <jralls> test-userdata-dir fails every test with a variant of C:\Users\John Ralls\AppData\Roaming\Gnucash oo vs. c:\gcdev64\msys2 mp\Gnucash oo . It does, however, link.
03:10:33 <gncbot> gjanssens: Sent 7 hours and 41 minutes ago: <jralls> The Win7 build dies at compiling unittest-support.go with a link error in libtest-core-guile.
03:26:02 <gjanssens> @tell jralls I see the same result when running test-userdata-dir so I can dive in to fix that. The good thing is the test runs without crashing (which it did before), so we're advancing step by step
03:26:02 <gncbot> gjanssens: The operation succeeded.
03:27:02 <gjanssens> @tell jralls When the build fails is gettext already compiled ? And on Win10 whe gettext fails is unittest-support already compiled?
03:27:02 <gncbot> gjanssens: The operation succeeded.
03:28:44 <gjanssens> @tell jralls Have you also tried simply continuing to rerun make (perhaps with -j 4) until the build stops compiling new targets ? I still wonder whether we're not seeing some obscure and indirect dependency not being built in time.
03:28:44 <gncbot> gjanssens: The operation succeeded.
03:29:30 *** mrklintscher19 has joined #gnucash
04:02:16 *** jotrago has quit IRC
04:02:49 *** fekepp has joined #gnucash
04:07:42 *** jotrago has joined #gnucash
04:32:27 *** fekepp has quit IRC
05:10:09 *** pilotauto_ has quit IRC
05:26:58 *** jotrago1 has joined #gnucash
05:27:59 *** jotrago has quit IRC
05:27:59 *** jotrago1 is now known as jotrago
05:33:25 *** fekepp has joined #gnucash
06:11:12 *** jotrago has quit IRC
06:20:54 *** storyjesse has joined #gnucash
06:51:26 *** Jimraehl1 has joined #gnucash
06:52:24 *** Jimraehl1 has left #gnucash
06:58:45 *** fabior has quit IRC
07:18:35 *** O01eg has joined #gnucash
07:35:40 *** User has joined #gnucash
07:56:40 *** rickoehn has joined #gnucash
08:23:17 <gjanssens> @tell jralls I have focussed until now on getting the test suite building. It does so now in my mingw32 shell. There are lots of test failures still though (44 out of 100 fail), but it's a start.
08:23:17 <gncbot> gjanssens: The operation succeeded.
08:25:41 <gjanssens> @tell jralls I suspect many of the issues are path related (like test-userdata-dir already revealed). I'll tackle those as time permits, starting with test-userdata-dir as that one seems to affect your build.
08:25:41 <gncbot> gjanssens: The operation succeeded.
08:25:48 <warlord> .
08:27:20 <gjanssens> @tell jralls My most recent build was with CMAKE_SYSROOT_DIR set to /d/gcdev64/msys2 rather than with CMAKE_FIND_ROOT_PATH. That did seem to build find so it looks like we may use that variable and not need to patch FindSWIG.cmake
08:27:20 <gncbot> gjanssens: The operation succeeded.
08:28:24 <warlord> gjanssens: did you try building the distribution package and ensuring that that succeeds?
08:28:38 <warlord> (on a system that doesn't have msys)?
08:31:58 <gjanssens> warlord: we're not that far yet. We first have to figure out why the build fails on jralls' systems and not on mine.
08:45:41 <warlord> True.
08:45:48 <warlord> And then figure out why the ovirt vm hangs.
08:49:08 *** jotrago has joined #gnucash
08:51:33 *** storyjesse has quit IRC
09:28:59 *** GorillaApe has joined #gnucash
09:29:07 <GorillaApe> Hello
09:30:18 *** fabior has joined #gnucash
09:36:13 *** Chris has joined #gnucash
09:36:17 <Chris> @tell warlord So I'm playing with scm and if I invoke (gncTaxTableGetTables (gnc-get-current-book)) this procedure returns #<swig-pointer GncInvoiceList * b8bbe50> -- I was expecting something like #<swig-pointer GList * b8bbe50>. Is this a bug?
09:36:17 <gncbot> Chris: The operation succeeded.
09:40:40 <warlord> .
09:40:40 <gncbot> warlord: Sent 4 minutes ago: <Chris> So I'm playing with scm and if I invoke (gncTaxTableGetTables (gnc-get-current-book)) this procedure returns #<swig-pointer GncInvoiceList * b8bbe50> -- I was expecting something like #<swig-pointer GList * b8bbe50>. Is this a bug?
09:42:47 <warlord> Chris: possibly... a GncInvoiceList is just a GList via typedef, so to *C* they are equivalent.
09:42:54 <warlord> The API (in C) returns a GList
09:43:07 <warlord> So if there is a bug it's in the swig wrapper around that function..
09:44:58 <Chris> Ok I'm digging through and am trying to enumerate the taxtables without understanding C, so far not so good :(
09:46:23 <warlord> Chris: well, you will need to understand C a bit. But I think it's a bug in the wrapper. We'll need to typedef a TaxTableList to GList (similar to the GncInvoiceList), and then in business-core.i you'll need to add a GLIST_HELPER_INOUT for the taxtablelist.. and then fix the API wrapper for the gncTaxTableGetTables.
09:46:44 <warlord> Just follow the recipe for how GncInvoiceList is used. Of GncTaxTableEntryList
09:47:34 <Chris> Thx for pointers! back to spelunking
09:49:10 <warlord> enjoy!
09:49:58 <Chris> :)
10:03:34 *** fabior has quit IRC
10:17:57 *** fabior has joined #gnucash
10:20:41 <Chris> :( definitely bug. I cannot enumerate taxtables via scm. I can enumerate gncTaxTableEntryList easily via (gncTaxTableGetEntries ...) which returns a list. But if I create several taxtables, (gncTaxTableGetTables (gnc-get-current-book)) always returns just one GncInvoiceList, which is not even a list. I wouldn't know how to fix C or write an appropriate bug report. Otherwise the API is very...
10:20:42 <Chris> ...well thought out.
10:21:33 <warlord> A GncInvoiceList should be a list of invoices..
10:21:58 <warlord> I don't know why it's returning that, tho. Most likely the type system is confused because of all the various typecasts of GList
10:22:12 <warlord> Chris: feel free to send in a PR with the patches to make it work.
10:25:07 <Chris> warlord: I would, except I really do not know C. I'll need to find another way.
10:27:10 <warlord> You don't really need to "know C".. Just copy and modify the existing pieces. It's a change to the header and a change to the swig interface file.
10:32:02 <Chris> :-o I'll look into it but this is a couple levels deeper than I expected to look at in this mission
10:34:01 *** kael has joined #gnucash
10:34:19 <warlord> Chris: it's really pretty straightforward...
10:52:30 *** fekepp has quit IRC
11:07:57 <lmat> GorillaApe: hi
11:20:52 *** ArtGravity has joined #gnucash
11:45:21 <jralls> .
11:45:21 <gncbot> jralls: Sent 8 hours and 19 minutes ago: <gjanssens> I see the same result when running test-userdata-dir so I can dive in to fix that. The good thing is the test runs without crashing (which it did before), so we're advancing step by step
11:45:22 <gncbot> jralls: Sent 8 hours and 18 minutes ago: <gjanssens> When the build fails is gettext already compiled ? And on Win10 whe gettext fails is unittest-support already compiled?
11:45:23 <gncbot> jralls: Sent 8 hours and 16 minutes ago: <gjanssens> Have you also tried simply continuing to rerun make (perhaps with -j 4) until the build stops compiling new targets ? I still wonder whether we're not seeing some obscure and indirect dependency not being built in time.
11:45:24 <gncbot> jralls: Sent 3 hours and 22 minutes ago: <gjanssens> I have focussed until now on getting the test suite building. It does so now in my mingw32 shell. There are lots of test failures still though (44 out of 100 fail), but it's a start.
11:45:25 <gncbot> jralls: Sent 3 hours and 19 minutes ago: <gjanssens> I suspect many of the issues are path related (like test-userdata-dir already revealed). I'll tackle those as time permits, starting with test-userdata-dir as that one seems to affect your build.
11:45:26 <gncbot> jralls: Sent 3 hours and 18 minutes ago: <gjanssens> My most recent build was with CMAKE_SYSROOT_DIR set to /d/gcdev64/msys2 rather than with CMAKE_FIND_ROOT_PATH. That did seem to build find so it looks like we may use that variable and not need to patch FindSWIG.cmake
12:00:51 *** Mechtilde has joined #gnucash
12:13:35 *** kael has quit IRC
12:26:36 *** mrklintscher110 has joined #gnucash
12:26:44 *** mrklintscher19 has quit IRC
13:22:35 <jralls> gjanssens: -DCMAKE_SYSROOT_DIR=/c/gcdev64/msys2 isn't working for me.
13:23:16 <jralls> gjanssens: You added it to the cmake options in the gnucash module, right?
13:41:11 <jralls> gjanssens: Ah, it's -DCMAKE_SYSROOT and NO_CMAKE_FIND_ROOT_PATH needs to be removed from FindSWIG.cmake for it to work.
13:43:38 *** Mechtilde has quit IRC
13:56:07 *** tuxd00d has joined #gnucash
13:58:38 <jralls> gjanssens: After looking at the unittest-support.go problem a bit I decided it looked like Windows had gotten messed up so I removed gcdev64 and did a system restore, then rebuilt everything from scratch. Now it compiles all the way through on Win7.
14:01:02 <jralls> gjanssens: So I'll add a a patch step for FindSWIG and then try that on Win10 just in case I've borked that environment somehow.
14:03:45 *** nomeata has joined #gnucash
14:07:01 *** mrklintscher110 has quit IRC
14:13:40 <warlord> Heh. Adonay just suggested we use Namazu for searching.. Hahaha
14:13:59 <warlord> My response is going to be "get it back into Fedora and I'll re-install it on the server"
14:16:39 <gjanssens> jralls: CMAKE_SYSROOT indeed. I see the same behaviour here indeed, although the find_path documentation is pretty ambiguous on this. I had interpreted it as if NO_CMAKE_FIND_ROOT_PATH was only influencing CMAKE_FIND_ROOT_PATH, but clearly that's wrong.
14:17:22 <jralls> Yeah, I had the same hope.
14:18:09 <gjanssens> Another approach could be to temporarily override CMAKE_INCLUDE_PATH in our CMakeLists.txt file ?
14:20:11 <gjanssens> Anyway, that can be fine-tuned later on if we care enough
14:20:34 <gjanssens> I'm glad you got it to build on Win7
14:26:46 *** frakturfreak has joined #gnucash
14:28:44 *** nomeata has quit IRC
14:28:57 <gjanssens> jralls: while you're changing the gnucash-modules file, I saw there's a hardcoded c:/gcdev64/msys2 in there for CMAKE_PREFIX, which won't work on my system. Can that be replaced with %(prefix)s ?
14:38:09 <jralls> gjanssens: I don't think that works anywhere except autogen-template and besides it points to c:/gcdev64/inst (or d:/gcdev64/inst on your system). But we can export an environment variable and use that. How does $ROOT_DIR sound?
14:39:05 <gjanssens> jralls: right. I misread the use of %(prefix)s
14:39:34 <gjanssens> $ROOT_DIR is fine with me
14:43:00 <gjanssens> jralls: or since there is already SRCROOT, why not MYS2ROOT ?
14:43:42 <gjanssens> I presume you will be exporting this from jhbuild and preconfigure jhbuild from setup_mingw64.ps1 ?
14:44:09 <jralls> I like MSYS2ROOT.
14:44:20 <gjanssens> Ok
14:45:09 <jralls> Yes, by way of jhbuildrc.in.
14:55:07 *** fabior has quit IRC
15:05:37 <gjanssens> jralls: saw your commits. $MSYS2ROOT got parsed by your shell :)
15:05:45 <gjanssens> In the comment that is
15:06:46 <jralls> Oh well. Nothing to do for it now. ;-(
15:08:24 <gjanssens> Nope
15:10:01 <jralls> I'm summoned for lunch, BIAB.
15:10:07 *** jralls is now known as jralls_afk
15:24:11 *** CDB-Away has joined #gnucash
16:16:22 *** jralls_afk is now known as jralls
16:25:27 <gjanssens> jralls: fyi I have also run make install and tried running gnucash
16:25:57 <gjanssens> It runs but apparently it ignores our precompiled scheme files. It compiles them all again
16:26:04 *** frakturfreak has quit IRC
16:26:45 <jralls> Did you set GUILE_COMPILED_PATH to point at them?
16:30:31 <gjanssens> The environment file should take care of that.
16:30:51 <gjanssens> I see that the autocompiled files get .scm.go as extension while our selfcompiled ones have .go
16:31:08 <gjanssens> I tried renaming our files but that doesn't change anything.
16:31:55 <gjanssens> I'll run in through gdb first to verify whether the environment is picked up correctly
16:36:24 *** fekepp has joined #gnucash
16:50:55 <gjanssens> jralls: turns out guile is pretty picky about path name notation :(
16:51:31 <gjanssens> D:/gcdev64/inst/... works, /D/gcdev64/inst doesn't, nor does D:\\gcdev\\inst\\...
16:54:43 <jralls> Yeah, I noticed. That's what 9d2a4b2ef0 was about.
16:57:46 <jralls> Well, this is weird: I recreated gcdev64 and tried to run jhbuild and got " 0 [main] python2 7408 child_info_fork::abort: address space needed by 'msys-ssl-1.0.0.dll' (0x6D0000) is already occupied"
17:04:21 *** fekepp has quit IRC
17:09:49 <jralls> The solution was to use a cygwin utility called "rebase" on the dll.
17:11:43 <gjanssens> Wow. And what caused this ? I hope other people don't need to do this...
17:14:43 <jralls> Heck if I know. It happened the first time I tried jhbuild. Might have something to do with an image left over from the previous incarnation, though it did survive a reboot.
17:18:47 <gjanssens> Ok We'll see
17:21:19 *** rofl has joined #gnucash
17:24:21 *** rofl has quit IRC
17:48:20 <gjanssens> jralls: I have just pushed a commit to gnucash-on-windows that should fix the googletest build and install errors
17:48:36 <jralls> Cool.
17:48:54 <gjanssens> One annoyance less :)
17:50:32 <gjanssens> Now I wish xmlsec would not interrupt each time because it refuses to work from a dirty directory
17:50:59 <gjanssens> That module gets rebuild each time jhbuild build is run, triggering a rebuild of aqbanking as well.
17:53:45 <jralls> Yes, I know. It's a tagged release on github so maybe if we build it from the tarball instead it would be nicer.
17:55:18 <gjanssens> What causes jhbuild to consider it dirty ?
18:01:23 <jralls> It's not jhbuild, it's git. The build is doing something so that when jhbuild tells git to checkout the specified revision it refuses.
18:02:02 <jralls> If we switch it to a tarball build then git won't get involved.
18:03:03 <gjanssens> Ok. Got it.
18:08:17 *** jotrago has quit IRC
18:18:06 <jralls> gjanssens: You can just rebuild gnucash by saying 'jhbuild -f /path/to/jhbuildrc buildone gnucash' or 'jhbuild -f... build --start-at=gnucash', though the latter will also build gnucash-docs.
18:18:32 <gjanssens> jralls: thanks, that's useful
18:18:52 <gjanssens> Btw, I've just pushed a minor fix for your MSYS2ROOT definition
18:19:16 <gjanssens> msys2 needed to be quoted as it's a litteral string
18:20:12 <gjanssens> I'm mentioning each commit I push primarily to let you remind you to pull before pushing your own commits.
18:29:51 *** jotrago has joined #gnucash
18:30:38 <jralls> I'd fixed that too, just hadn't pushed it yet.
18:30:56 <jralls> Don't worry about that, git will tell me!
18:32:01 <gjanssens> Right :)
18:32:22 *** dollabills has joined #gnucash
18:35:47 <dollabills> RAKING IN THE DOLLA DOLLA BILLS
18:35:47 <dollabills> RAKING IN THE DOLLA DOLLA BILLS
18:35:47 <dollabills> DONATIONS ARE NEEDED FOR THIS GREAT WORK!!
18:35:47 <dollabills> PLEASE VISIT #WEECHAT
18:35:47 <dollabills> IRC.FREENODE.NET FOR MORE INFORMATION!!
18:35:47 <dollabills> type !donation in channel for donation info
18:35:47 <dollabills> jotrago CDB-Away tuxd00d ArtGravity Chris GorillaApe rickoehn User O01eg gjanssens bertbob Cuare jralls entreprnr To7 wget fiddlerwoaroof karelk Artefact2 weasel Mobtrel Simon warlord lmat CDB-Man chf mikee icasdri GabrieleV gncbot g5pw Cork jonas jethrogb Unhammer flips yuriks puck akv Derperperd
18:35:47 *** dollabills has left #gnucash
18:40:15 *** pilotauto has joined #gnucash
18:47:43 *** dollabills has joined #gnucash
18:51:54 *** dollabills has quit IRC
18:55:41 <jralls> gjanssens: Well, I switched xmlsec to tarball. It still insists on rebuilding every time but at least it doesn't stop for the dirty tree thing.
18:56:18 <gjanssens> jralls: at the end of the install phase for gnucash I get this error now: https://pastebin.com/c5BuaCiy
18:56:56 <gjanssens> jralls: tx for the xmlsec improvement
18:58:31 *** rickoehn has quit IRC
18:59:42 <jralls> That's because the jhbuild patch didn't apply. I've got a fix, just need to push it.
18:59:58 <jralls> Woo-hoo, gnucash built all the way through on Win10.
19:00:39 <gjanssens> Yay! Mexican wave! \o/
19:48:47 <warlord> Yay!!
19:50:17 <gjanssens> jralls: It's my time to go to bed (way over actually but hey, time flies when you're having fun)
19:51:03 <gjanssens> I was working on fixing the guile autocompilation due to bad paths, but that needs some more debugging
19:51:03 <jralls> You must have been having a lot of fun! Sleep well!
19:51:44 <gjanssens> glib's g_build_filename is not doing as advertised (that is reusing the last used separator)
19:51:51 <jralls> OK. There are still a few warts in jhbuild, like it isn't always updating the packagedb; I think that's why it doesn't recognize that xmlsec doesn't need to rebuild.
19:51:55 <gjanssens> I'll continue that tomorrow. It's not blocking further work
19:52:23 <jralls> Right. The next step is to test the packaging script and then test on warlord's VM.
19:52:24 <gjanssens> Is there a step to build the installer already ?
19:52:45 <jralls> It's another powershell script.
19:52:45 <gjanssens> There, we were thinking about the same thing :)
19:52:50 <gjanssens> Ok
19:52:58 <gjanssens> So see you tomorrow
19:53:06 <jralls> Goodnight.
19:54:02 *** gjanssens has quit IRC
20:16:24 *** Chris has quit IRC
21:10:42 <jralls> warlord: Now that the Gnome folks have installed nickserv and chanserv is there any need to camp on the channel?
21:14:59 *** User has quit IRC
22:35:17 *** ArtGravity has quit IRC
22:41:50 *** kael has joined #gnucash
22:52:54 *** lmat has quit IRC
22:54:09 *** lmat has joined #gnucash
23:40:36 *** kael has quit IRC