| rinigus | mal: idea was to have a fixed target and release first. we have scripts ready to bump Qt and KF versions as well as change build targets. so, should be simple when we get there | 06:47 |
|---|---|---|
| rinigus | nephros: part of the reply to your questions is in https://github.com/sailfishos-chum/qt6/blob/main/updates.md . we also have applications.obs and maybe they go there | 06:48 |
| rinigus | if we have several packages with 24.xx.yy it would make sense to add them. .kde looks good option. then we would have to add handling of these in the scripts | 06:50 |
| rinigus | piggz: when comparing angelfish requirements with fedora, looks like qcoro is missing in sfos spec. is it the latest version? | 07:07 |
| rinigus | ... in github? | 07:07 |
| piggz[m] | Rinigus: qcoro gets pulled in by futuresql anyway | 07:50 |
| rinigus | piggz: thanks! | 07:51 |
| rinigus | piggz: I have added a build for sfos keyboard at chum:testing | 07:57 |
| rinigus | dealing with qt5compat prebuild | 07:57 |
| piggz[m] | rinigus: yeah, i had to build that locally ... im a little confused how something work locally but not on OBS ... supersonik also fails on OBS, but works in SB2 locally | 08:11 |
| rinigus | piggz: I think this sb2 bug has cost us a lot of time. very frustrating... | 08:14 |
| rinigus | hopefully it will be fixed soon | 08:14 |
| piggz[m] | rinigus: also, the qcoro i have used is a different GH url to the one in fedora, but -seems- more official | 08:17 |
| rinigus | piggz: but qcoro and futuresql we can add to obs directly, right? no need to have the prebuilt | 08:18 |
| piggz[m] | yes i think so | 08:19 |
| piggz[m] | they build fine on obs in my qt6apps project | 08:19 |
| rinigus | piggz: will you add them? or should I do it? would be nice to tag them as well | 08:22 |
| piggz[m] | i will tag and submit to testing | 08:22 |
| rinigus | qt5compat is uploaded, should be soon available | 08:27 |
| piggz[m] | rinigus: those 2 are building in testing 4.6 | 08:36 |
| rinigus | piggz: excellent! I will add qcoro and future to our lists in qt6. that way we will keep the updated | 08:36 |
| rinigus | piggz: when building locally, I have it failing exactly as at obs (complaining about qtcore/... folder) | 08:45 |
| piggz[m] | rinigus: which package? | 08:46 |
| rinigus | angelfish | 08:46 |
| rinigus | as in https://build.sailfishos.org/package/show/home:piggz:qt6apps/angelfish | 08:46 |
| piggz[m] | that looks like a missing dep... | 08:47 |
| rinigus | hmm private devel missing? | 08:47 |
| rinigus | let's try | 08:48 |
| piggz[m] | yeah, qt6-qtbase-private-devel | 08:48 |
| rinigus | piggz: yes, that's the one. | 08:49 |
| piggz[m] | thats installed in my sfdk, so was available in the build | 08:49 |
| rinigus | do I push my spec changes to kf6 branch? | 08:50 |
| piggz[m] | sure | 08:50 |
| piggz[m] | rinigus: for tokodon, i need qtkeychain, which nephros has been working on | 08:56 |
| rinigus | its in chum:testing already :) | 08:56 |
| rinigus | as it builds nicely, without prebuilt | 08:56 |
| rinigus | piggz: ^ | 08:56 |
| piggz[m] | gr8 | 08:56 |
| rinigus | piggz: kf6 branch of angelfish is updated | 09:02 |
| rinigus | would you trigger new build at your repo? | 09:02 |
| rinigus | (and maybe remove qcoro futuresql and sfos-kbd from there as they are at chum:testing as well) | 09:03 |
| piggz[m] | k | 09:03 |
| piggz[m] | nearlt got tokodon to build | 09:03 |
| rinigus | piggz: and now we have it again: https://build.sailfishos.org/package/show/home:piggz:qt6apps/angelfish | 09:12 |
| piggz[m] | yeah, qmlcachegen like in supersonik... | 09:13 |
| rinigus | I do wonder how far is jolla to that sb2 fix. it seems like a lot of time and hacks is going to fight it | 09:14 |
| piggz[m] | i wonder if qmlcachegen can be disabled at build time, and have it make the cache on run | 09:15 |
| piggz[m] | but why does it work locally, grrr | 09:15 |
| rinigus | +1 hack :( | 09:15 |
| rinigus | it could be related to vm handling as was pointed out somewhere | 09:16 |
| piggz[m] | https://doc.qt.io/qt-6/cmake-source-file-property-qt-qml-skip-cachegen.html ,,,, i dont think we want do do this though | 09:16 |
| piggz[m] | fun | 09:17 |
| * piggz[m] uploaded an image: (108KiB) < https://matrix.org/oftc/media/v1/media/download/AV3PDbO-bKjWwdrGJIRLL4g7dwy9iRbARuMTdNXJ2oBQElEHWzK0s3ISmY3Gk9uTFbLUYQspAeHMqWCyh-OT2GJCeS7CpbWQAG1hdHJpeC5vcmcvR2FiSXRxZHBIZXV2alRKY3puWnV3UXhR > | 09:17 | |
| rinigus | nope, looks like qmlcache is good. maybe we should wait till the bug fix is done? and test meanwhile with local app builds? | 09:18 |
| piggz[m] | yeah | 09:18 |
| rinigus | as soon as the bugfix is ready we would be able to continue there. if we see that it takes tooo looong, maybe we should setup additional obs for qt6. with native builds | 09:21 |
| rinigus | but again, its a lot of time that will go down the drain as soon as sb2 is fixed | 09:22 |
| nephros | I don't have an SDK handy, but just for kicks, can one of you try building a failing one with this? | 09:24 |
| nephros | https://build.sailfishos.org/package/binaries/home:nephros:devel:qt6:devel/sb2plus?repository=4.6_i486 | 09:24 |
| nephros | rinigus: agreed, but OTOH the qt6 stuff is so much nicer to package, without all the /opt shenanigans! | 09:27 |
| rinigus | nephros: I don't have sdk either - using my native build env instead. so no sb2 here... | 09:28 |
| nephros | ah, ok. well, it's jollas task to fix anyway. But they are busy taping C2 shipping boxes. | 09:30 |
| rinigus | nephros: its way better to package, indeed! I think most of the time (75%-90% ?) we have spent fighting this bug. either piggz trying to tip-toe around it in qtbase or me writing tbuilder. | 09:30 |
| rinigus | well, we don't mind to help to fix it, just seems like we don't have ways to do it ... so, I do hope they work on it | 09:32 |
| nephros | With non-qt6 packages, I'm often getting: | 09:41 |
| nephros | [ 29s] cc1plus: warning: /usr/include/qt6/QtCore/6.7.2: No such file or directory [-Wmissing-include-dirs] | 09:41 |
| nephros | [ 29s] cc1plus: warning: /usr/include/qt6/QtCore/6.7.2/QtCore: No such file or directory [-Wmissing-include-dirs] | 09:41 |
| nephros | ... is there something wrong with the qtbase include packaging? | 09:41 |
| nephros | may hidden by requiring the private headers? | 09:42 |
| rinigus | nephros: something could be wrong, but I don't know what. getting same messages | 09:45 |
| rinigus | piggz: indeed, angelfish is rather crashy | 09:45 |
| piggz[m] | indeed :) | 09:48 |
| piggz[m] | maybe something in the console? | 09:48 |
| rinigus | piggz: `corrupted double-linked list` was just printed when running without qt-runner | 09:57 |
| rinigus | messages with getInjectedScript also, but those should be ok | 09:57 |
| poetaster | piggz[m], rinigus I noticed in the qt5.15 angelfish builds that the autogen qmlcache stuff is not present. | 11:23 |
| poetaster | I was thinking that cachegen is not enabled on the older builds? | 11:23 |
| rinigus | piggz: getting more of those messages with corrupted lists. this time with qt-runner as well. | 11:39 |
| rinigus | poetaster: piggz found an option on how to disable qmlcachegen (see above). in principle we can try to compile with that and see what will happen | 11:40 |
| rinigus | if it works at obs | 11:40 |
| nephros | rinigus: Probably not related, I get the corrupted lists almost every time I start retroarch. There it has something to do with OpenGL/EGL and/or Wayland initialization. | 11:43 |
| rinigus | nephros: that happens usually after a bit of time. not sure if initialization is involved. I have a feeling that we will need to make maybe some simple browser and see if the issue comes from webengine or angelfish code | 11:45 |
| rinigus | ... if someone feels like debugging this huge code. or try to bump all to latest qt first | 11:47 |
| rinigus | I'll try to see whether disabling cachegen would allow us to stay at obs and experiment there | 11:47 |
| poetaster | rinigus, that'd be great. | 11:48 |
| rinigus | just beware that home:rinigus:qt6 will be broken now :) | 11:48 |
| rinigus | poetaster: that failed - no such cmake option that I can set from cmake config as far as I can see | 12:56 |
| poetaster | hmmm. odd. | 13:19 |
| poetaster | I think the target properties have to be set first before you can apply that directive. https://doc.qt.io/qt-6/qt-target-qml-sources.html#source-file-properties | 13:49 |
| poetaster | rinigus, where are you building? | 13:51 |
| poetaster | I'm no cmake guy, but looking at https://invent.kde.org/network/angelfish/-/blob/master/CMakeLists.txt?ref_type=heads there is no appropriate target to execute that cachegen directive for. | 13:57 |
| poetaster | ah, kde. not made here syndrome. | 13:59 |
| rinigus | poetaster: that was at obs and I disabled those now to avoid pointless builds: home:rinigus:qt6 | 14:05 |
| poetaster | ok. I just wanted to see Cmake lists. | 14:06 |
| poetaster | There might be a way aroudn https://doc.qt.io/qt-6/qt-target-qml-sources.html#source-file-properties but I'll have to dig through all that kde foo. | 14:07 |
| rinigus | poetaster: no need, you could start with qtdeclarative. it fails already there with sb2 | 14:07 |
| poetaster | I was at https://build.sailfishos.org/build/home:piggz:qt6apps/sailfish_latest_aarch64/aarch64/angelfish/_log | 14:08 |
| poetaster | there I don't see more than the one fatal? | 14:09 |
| poetaster | ECMFindQmlModule and ECMQmlModule is where I'm at | 14:10 |
| rinigus | I don't follow, sorry | 14:10 |
| rinigus | that build failed | 14:11 |
| poetaster | Ah, I'm trying to figure out how kde manages qml module imports and the cmake macros it uses. since the cmake file they provide does not 'show' the qt targets we need. | 14:11 |
| rinigus | but if we cannot properly build already qtdeclarative then angelfish is already waay too far :) | 14:12 |
| poetaster | sorry, you mean the qtdeclarative package, correct? | 14:13 |
| rinigus | right | 14:13 |
| poetaster | ok, thanks. I'll go read some logs. | 14:13 |
| rinigus | those could be gone though. maybe piggz has them somewhere around | 14:14 |
| poetaster | https://build.sailfishos.org/package/show/home:piggz:qt6/qt6-qtdeclarative | 14:14 |
| rinigus | Aarch64 disabled. You would have to build yourself then it finds some smaller package to test with | 14:16 |
| poetaster | rinigus, but that's the same error. the qmlcompiler. | 14:16 |
| poetaster | there is a build log showing the fail. | 14:16 |
| rinigus | Then it's good if the log is still there | 14:17 |
| poetaster | https://build.sailfishos.org/build/home:piggz:qt6/sailfish_latest_aarch64/aarch64/qt6-qtdeclarative/_log | 14:17 |
| poetaster | kde in their infinite wisdom duplicate qt functions everywhere. The remnanats where we are concerned: https://api.kde.org/ecm/module/ECMQmlModule.html | 14:18 |
| poetaster | ah, whatever. it's macro on top of macro hell. I'm backing out ... | 14:31 |
| poetaster | oh, https://invent.kde.org/network/angelfish/-/blob/master/src/settings/CMakeLists.txt?ref_type=heads one could try adding NO_CACHEGEN to ecm_add_qml_module( | 14:35 |
| poetaster | shit. one would have to add that to each module declaration. sigh. | 14:38 |
| rinigus | poetaster: yes, I have the same impression. which is way too much | 14:41 |
| poetaster | Yeah, interesting as an excercies to a point. My guess. the ecm_add_qml_module foo is the issue, since it can/does circumvent how Qt deals with file paths for resources. | 14:41 |
| poetaster | But I am neither a Qt, KDE or Cmake specialist :) | 14:42 |
| poetaster | interesting in the angelfish case is that the error appears for the settings sub dir which is defined as a module using the kde macro. But I'm guessing. | 14:43 |
| poetaster | ah, it looks like they only did the settings as a module. As in the CMakesLists | 14:52 |
| poetaster | above | 14:52 |
| piggz[m] | what have i missed in qt6 land with my day outside?? | 17:37 |
| rinigus | piggz: not super-much. no simple way was found to disable qmlcachegen. angelfish crashes with the same message, but sometimes message is missing. and crashing surely too much | 18:07 |
| piggz[m] | rinigus: how does the demo browser behave? | 18:08 |
| rinigus | don't have it around - it wasn't very good. but yeah, we would have to start separating angelfish from webengine | 18:09 |
| rinigus | also, maybe qt bump would help | 18:09 |
| rinigus | piggz: I'd like to make small version bumps in pure maps stack first. kind of outside qt6 land for a bit :) | 18:12 |
| piggz[m] | rinigus: the new tile provider? | 18:13 |
| rinigus | yes that as well . also maplibre qt bindings got finally tagged and would make sense to bump to new api | 18:14 |
| rinigus | piggz: was cleaning today and looks like removed nanobrowser folder too :( | 18:16 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!