2019-06-08 GnuCash IRC logs
01:02:27 *** tienne has joined #gnucash
01:18:07 *** fell_laptop has joined #gnucash
01:18:08 *** ChanServ sets mode: +o fell_laptop
01:20:29 *** fell has quit IRC
01:27:27 *** Yotson has quit IRC
01:38:25 *** Mechtilde has joined #gnucash
01:39:33 *** tienne has quit IRC
01:43:27 *** Yotson has joined #gnucash
01:48:45 *** Mechtilde has quit IRC
01:49:41 <chris> @tell gjanssens it would be fine to start deleting unused modules such as report-collectors.scm in master right?
01:49:41 <gncbot> chris: The operation succeeded.
01:56:48 *** gjanssens has joined #gnucash
01:56:48 *** gncbot sets mode: +o gjanssens
01:56:48 *** ChanServ sets mode: +o gjanssens
01:56:55 <gjanssens> .
01:56:55 <gncbot> gjanssens: Sent 7 hours and 48 minutes ago: <jralls> warlord the new guile is hanging up on the Windows VM, you can see it in Task Manager. Maybe I should roll back the base commit to the April one.
01:56:56 <gncbot> gjanssens: Sent 6 hours and 3 minutes ago: <jralls> A clean build on Win7 with gcc-9.1 works for me with last Monday's 170691 but fails with today's HEAD. Failure is in building qif-import.go, problem is inability to load libgnc-gnome.dll. I've started a bisect but probably won't get it finished tonight.
01:56:57 <gncbot> gjanssens: Sent 5 hours and 54 minutes ago: <chris> - the (use-modules (...)) is a macro. I suspect we could redefine the macro and attach a backward-compatibility list.
01:56:58 <gncbot> gjanssens: Sent 7 minutes ago: <chris> it would be fine to start deleting unused modules such as report-collectors.scm in master right?
01:59:20 <gjanssens> chris: yes we can start deleting those. I will do this on my branch as I'm currently working in that area.
02:00:41 <gjanssens> Redefining the module is a clever hack. Something you like to/can experiment with ?
02:58:41 *** gncbot has joined #gnucash
03:46:54 *** Mechtilde has joined #gnucash
03:59:38 *** Mechtilde has quit IRC
05:14:33 *** gncbot has joined #gnucash
05:16:10 <gjanssens> From there I can only guess - I removed both libgnc-gnome.dll and libgncmod-gnome-utils.dll from the inst directory which allowed the build to complete
05:17:16 <gjanssens> I'll retry to see which of the two is really causing the build failure
05:32:16 <gjanssens> Further test: only removeing libgncmod-gnome-utils from inst/bin is sufficient (which is kind of what I was expecting)
05:39:23 *** Mechtilde has joined #gnucash
05:44:44 *** Mechtilde has quit IRC
05:45:15 *** Mechtilde has joined #gnucash
05:51:39 <gjanssens> But that's as far as I get. Dependency walker does seem to pick up the right one if I preset the PATH variable to what the build rule for qif-import.dll sets it to (well, I had to convert it to a CMD style SET PATH="" form to work)
05:51:59 <gjanssens> I don't really know how to take it from there.
05:52:58 <gjanssens> Oh and interestingly I don't see the guile hangs on my local system
05:55:31 *** storyjesse has quit IRC
07:03:06 *** Mechtilde has quit IRC
07:14:31 *** oozer has joined #gnucash
07:24:46 *** warlord has joined #gnucash
07:25:12 *** ChanServ sets mode: +qo warlord warlord
07:25:14 <warlord> .
07:25:17 *** warlord sets mode: +o gncbot
07:27:36 <warlord> Looks like my "hosting facility" had some power hits overnight. Also looks like everything came back up successfully.
07:32:04 <chris> gjanssens (use-modules ...) calls (resolve-interface ...) internally via complicated macros
07:33:39 *** Mechtilde has joined #gnucash
07:40:21 *** Jimraehl1 has joined #gnucash
07:40:27 <gjanssens> chris: is there a way to redefine a macro ?
07:40:48 <gjanssens> Something like (redefine use-modules as official-use-modules)
07:41:49 <gjanssens> (define-macro use-modules (do-deprecation-dance modulename->adjusted-module-name)(officical-use-modules (adjusted-module-name))
07:41:52 *** Jimraehl1 has left #gnucash
07:42:10 <gjanssens> That's hyper simplified code of course, not syntax checked
07:44:14 <gjanssens> If we could inject such a snippet of code early in the startup, all calls to use-modules would first check if the requested module is deprecated, issue a warning and adjust the module name to the replacement module before calling the built-in use-modules (renamed to official-use-modules) macro
07:46:01 <chris> not sure i'm asking #guile
07:52:52 *** Mechtilde has quit IRC
08:06:54 *** Mechtilde has joined #gnucash
08:09:56 <gjanssens> ok
08:10:17 <gjanssens> BTW the deprecated scheme functions have been dropped on my branch :)
08:18:38 <chris> cool, breathing space!
08:19:02 *** User_ has joined #gnucash
08:19:33 <chris> report.scm will be completely transformed, especially if book-reports is suitable for merging - the latter needs a compatibility path for tabbed-reports loading
08:23:22 *** boldstripe has joined #gnucash
08:26:58 *** Mechtilde has quit IRC
08:32:53 *** Mechtilde has joined #gnucash
08:38:06 *** User_ has quit IRC
08:40:29 *** oozer has quit IRC
08:48:19 *** oozer has joined #gnucash
09:00:36 *** Mechtilde has quit IRC
09:02:17 *** Mechtilde has joined #gnucash
09:32:06 <gjanssens> chris: just a heads-up, in my local work report.scm will be renamed to report-core.scm as I want to use the report.scm name for the report module.
09:33:27 <gjanssens> you can check my work so far over here: https://github.com/gjanssens/gnucash/commits/master
09:33:51 <gjanssens> chris, jralls: unfortunately our master branch fails make check on ubuntu-16.04
09:34:28 <gjanssens> https://travis-ci.org/gjanssens/gnucash/jobs/543131707
09:34:48 <gjanssens> The timezone test bails out. Looks like something is not right on that platform
09:35:11 <gjanssens> I suspect libboost (which is 1.58 on Ubuntu 16.04)
09:40:50 <chris> I guess jqplot modules can be deprecated for 4.x series
09:41:14 <chris> that's a lot of scheme cleanup!
09:42:13 *** fell_laptop has quit IRC
09:43:10 <chris> I'm fine with report-core.scm -- no plans to tinker with it
09:47:09 *** jervin has joined #gnucash
10:04:53 *** Mechtilde has quit IRC
10:07:53 <chris> gjanssens: libboost failure may effect guile's utf8 handling too... test-html-utilities barfs on emoji string-sanitize
10:19:09 <gjanssens> chris: is there still code using jqplot (like some fallback code or something similar) ?
10:21:08 *** jervin has quit IRC
10:22:25 <chris> I don't think so
10:23:36 <gjanssens> Ok
10:34:28 <chris> Actually test-graphing.scm hasn't been converted to chartjs... perhaps I'll do it soon
10:34:47 *** boldstripe has quit IRC
10:36:39 <jralls> gjanssens: I don't have an installed libgnc-anything and I'm starting every build with an empty build directory. Reverting cc12fcf32 does fix the build for me, I'll analyze what's going on after.
10:37:16 <gjanssens> jralls: ok
10:39:20 *** boldstripe has joined #gnucash
10:43:59 <jralls> gjanssens: Travis is using Ubuntu 14.04 not 16.04, but that shouldn't have anything to do with the failure. test-html-utilities-srfi64 will have failed because Ubuntu-14.04 doesn't have a new enough Guile. test-gnc-timezone seems to be failing to get the timezone changed. That will take more investigation.
10:46:30 <gjanssens> jralls: no no, I was testing an updated travis script for master, bumping the ubuntu docker image to 16.04
10:46:46 <gjanssens> I did the same on a local docker image, got the same result
10:54:35 <gjanssens> Hmm maybe the problem is really that not all timezone info is installed, haven't tested that yet
10:56:56 *** Mechtilde has joined #gnucash
11:03:28 <gjanssens> That doesn't seem to be the problem: /usr/share/zoneinfo/America/Los_Angeles does exist
11:08:54 *** fell_laptop has joined #gnucash
11:08:54 *** ChanServ sets mode: +o fell_laptop
11:12:36 *** jervin has joined #gnucash
11:26:34 <chris> test-graphing.scm is not included in CMakeLists so it's effectively dead code. we can either delete the file, or upgrade to chartjs. then jqplot-be-gone
11:39:33 *** User_ has joined #gnucash
11:42:37 *** User_ has quit IRC
11:51:18 *** jervin has joined #gnucash
11:52:22 *** jervin has quit IRC
12:14:12 <gjanssens> chris: I think it may be useful to revive test-graphing.scm and add it to the Examples menu
12:14:42 <gjanssens> It doesn't have much other logic than to add charts so it's an easy one to learn from
12:15:02 <gjanssens> Of course it should be converted to chartjs for that to work...
12:32:53 <jralls> gjanssens: The job log that you linked to says it's Ubuntu 14.04.
12:36:14 <gjanssens> jralls: it says before the docker container is started
12:37:03 <gjanssens> But the build and test run are in a docker 16.04 container (using gcc 5.4)
12:39:24 *** jervin has joined #gnucash
12:43:15 <jralls> OIC, I didn't look far enough down.
12:45:25 <gjanssens> Np
12:45:44 <gjanssens> Got to leave early today
12:47:25 <jralls> OK.
12:47:50 <gjanssens> fyi if you want to experiment with my ubuntu 16.0 travis work, I have pushed it to a separate branch for now
12:47:52 <gjanssens> https://travis-ci.org/gjanssens/gnucash/builds/543182786
12:48:02 <gjanssens> https://github.com/gjanssens/gnucash/tree/travis1604
12:48:35 <gjanssens> That allowed me to continue my guile installation restructuring independently on https://github.com/gjanssens/gnucash/tree/master
12:49:30 <jralls> OK. I'll poke at it on Debian stable and see if it works there, and if we're going to use 16.04 instead of 18.04 as the LTS for v4.x I'll need a VM for troubleshooting.
12:50:29 *** Mechtilde has quit IRC
12:51:40 <gjanssens> I went for 16.04 as it's a currently still supported LTS
12:51:50 *** Mechtilde has joined #gnucash
12:51:59 <gjanssens> Would you rather skip it ?
12:54:32 <jralls> Based on your formula for Fedora it would seem to be the right choice for 4.x give a 2-year major cycle for GC.
12:54:58 *** Mechtilde has quit IRC
12:55:12 *** Mechtilde has joined #gnucash
12:57:49 <jralls> gcc5.4 supports all of C++14 and none of C++17; we need gcc7 for C++17. There's some nice constexpr stuff in C++17 but I guess we can wait a couple more years.
13:01:11 <gjanssens> :)
13:02:35 * gjanssens is happy the travis run was successful for the guile directory restructuring \o/ https://travis-ci.org/gjanssens/gnucash/builds/543182443
13:02:45 <gjanssens> A few more tweaks and then it's ready for master
13:02:49 <gjanssens> But got to go now.
13:02:54 <gjanssens> Till later
13:02:58 <jralls> Bye
13:03:31 *** gjanssens has quit IRC
13:05:35 *** Mechtilde has quit IRC
13:18:22 *** Mechtilde has joined #gnucash
13:18:36 *** lmat has quit IRC
13:26:21 *** Mechtilde has quit IRC
13:26:49 *** fell_laptop has quit IRC
13:26:51 *** fell has joined #gnucash
13:26:51 *** ChanServ sets mode: +o fell
14:12:19 *** oozer has quit IRC
15:01:24 *** oozer has joined #gnucash
15:02:48 *** warlord has quit IRC
15:03:05 *** warlord has joined #gnucash
15:03:44 *** Mechtilde has joined #gnucash
15:17:40 *** jervin has joined #gnucash
15:23:58 *** badger92 has quit IRC
16:03:10 *** boldstripe has quit IRC
16:05:20 *** Cork has quit IRC
16:14:41 *** Cork has joined #gnucash
16:25:05 *** jervin has quit IRC
16:26:36 *** Mechtilde has quit IRC
16:36:59 <jralls> chris: On Debian Stable (guile-2.0.13) and Ubuntu 16.04 (guile 2.0.x) test-utilities-srfi64 fails on the two emoji tests. Should those tests be done only on guile 2.2?
16:41:34 <jralls> chris: s/test-utilities-srfi64/test-html-utilities-srfi64/
17:03:06 *** Mechtilde has joined #gnucash
17:06:17 *** Mechtilde has quit IRC
17:28:17 *** Cork has quit IRC
17:28:37 *** lmat has joined #gnucash
17:30:07 *** Cork has joined #gnucash
18:16:06 *** guak has joined #gnucash
18:38:34 <jralls> @tell gjanssens I discovered while testing your travis1604 branch that you're putting some go files into lib/gnucash/scm/ccache/2.2/gnucash/engine/include. Let's not call it that, that name is commonly used for C header files.
18:38:34 <gncbot> jralls: The operation succeeded.
18:42:10 <jralls> @tell gjanssens I also found that your changes have broken `ninja check` unless `ninja` is run first. That actually starts with chris's borrowed/guile-json commit, but building engine-utilities.go fails earlier in HEAD on that branch.
18:42:10 <gncbot> jralls: The operation succeeded.
18:43:13 <jralls> @tell gjanssens The timezone problems didn't reproduce on Debian stable, so I'm setting up an Ubuntu VM to test there.
18:43:13 <gncbot> jralls: The operation succeeded.
19:06:57 *** warlord has quit IRC
19:41:06 *** warlord has joined #gnucash
19:50:35 *** guak has quit IRC
19:56:54 <jralls> @tell gjanssens I set up an Ubuntu 16.04 VM and provisioned it (after updating) using the commands from your ubuntu-16.04-docker, then built and tested your travis1604 branch. The emoji tests failed as expected but test-gnc-timezone passed. Hmm, one difference, I used the System Settings to add the deb-src repositories to sources.list and it didn't uncomment every deb-src line.
19:56:54 <gncbot> jralls: The operation succeeded.
20:02:24 <jralls> @tell gjanssens Removing all of the commented deb-src lines gets me a bunch of warnings about duplicates but no new updates.
20:02:24 <gncbot> jralls: The operation succeeded.
20:10:14 *** boldstripe has joined #gnucash
20:40:41 <jralls> @tell gjanssens As for Windows, I found I did in fact have libgnc-gnome.dll on the install path, once I deleted that gnucash builds cleanly. I've run pacman -Syuu on the build vm until it's clean, we'll see what happens with tomorrow's builds.
20:40:41 <gncbot> jralls: The operation succeeded.
20:51:45 <jralls> @tell gjanssens As for Windows, I found I did in fact have libgnc-gnome.dll on the install path, once I deleted that gnucash builds cleanly. I've run pacman -Syuu on the build vm until it's clean, we'll see what happens with tomorrow's builds.
20:51:45 <gncbot> jralls: The operation succeeded.
20:56:02 *** oozer has quit IRC
22:24:42 *** boldstripe has quit IRC
22:43:40 <chris> jralls: I guess guile-2.0.13 has SRFI64 but guile-2.0 still cannot handle non-UTF8 files. so agree test should probably be enabled only for guile-2.2 onwards.
22:46:09 <chris> gjanssens: I'll supercharge test-graphing.scm to demonstrate most exposed chartjs functionality, and a couple others to demonstrate how to modify arbitrary chartjs options tree... and numerous comments too
22:46:16 <chris> @tell gjanssens I'll supercharge test-graphing.scm to demonstrate most exposed chartjs functionality, and a couple others to demonstrate how to modify arbitrary chartjs options tree... and numerous comments too
22:46:16 <gncbot> chris: The operation succeeded.
22:52:01 *** boldstripe has joined #gnucash
22:53:17 *** jervin has joined #gnucash
22:53:19 <chris> @tell gjanssens jcowan says guile macros are not first class, so, cannot call old one
22:53:19 <gncbot> chris: The operation succeeded.
23:00:35 <chris> @tell gjanssens the solution is to copy defition of use-modules from guile sources and augment it. (I'd vote to skip this to avoid clutter...)
23:00:35 <gncbot> chris: The operation succeeded.
23:01:39 *** jervin has quit IRC