Sunday, 2022-01-02

piggzrinigus: so, currently, just need to drop the as_const changes?14:17
bionade24Nico: Can please explain me how I make application profiling with perf? I already recorded perf data but it doesn't seem to subsitute the adresses with the human readable symbols.14:34
riniguspiggz: yes, I think so. let's keep the rest14:58
riniguspiggz: could you make that as a separate commit with dropping as_const and then maybe it is time to reformat the code? I think it would nicely fit your PR14:58
riniguspiggz: together with reformatting, let's also update TS and I think we could make a release after that. would that be OK?15:03
piggzrinigus: you mean seperate commit as, remove the as_const changes locall and commit, or make them never appear in the PR?15:12
riniguspiggz: I think it is fair enough to add removal commit - so we would have one as suggested by clazy and the next after that correcting it15:13
piggzrinigus: pushed15:16
riniguspiggz: reading15:16
piggzrinigus: i also need to remove the include15:18
piggzrinigus: oh, also you thought the qlatin1string changere were not needed?15:19
riniguspiggz: I don't its needed, but I have no major point for or against it. so, let's keep it as it is in your commit now15:20
rinigusstarted rebuild at OBS15:20
rinigusfor gui, to check whether older sfos would work15:21
riniguspiggz: looks to be building fine for older sfos versions. do you want to reformat whole project? or is it too much...15:24
piggz[m]rinigus: how about reformat after merge15:25
piggz[m]rinigus: pushed another small change to match the clazy suggestion for QLatin1String15:25
riniguspiggz: sure, reformatting can be separate as well15:26
riniguspiggz: weird suggestion, but I guess they had a reason for it15:27
riniguslet's merge15:27
piggz[m]rinigus: cool.... i cant claim to understand all the clazy suggestions either!15:28
piggz[m]there were 3 left over that i couldnt figure out a way to resolve15:28
riniguspiggz: reading the end of - we could have used plain QString() there :) . but that we can do in future15:31
piggzyeah, i wondered that too15:31
piggznot that it makes a noticable difference!15:31
riniguspiggz: started reading reformat PR15:40
piggz[m]rinigus: cool ... shouldnt be much to look at though, just white space!15:41
riniguspiggz: merged.15:50
riniguspiggz:  I will regenerate TS and then push it directly to main branch15:51
riniguspiggz: TS strings updated; polish incorporated into builds15:59
rinigusall in main branch already15:59
piggzrinigus: another Pr incoming :)15:59
riniguspiggz: was going to suggest the releasing. well, you let me know when you think it is ready for it16:00
piggzrinigus: its quite minor...16:00
riniguspiggz: no problem - keep them coming. I am working on adding metadata to chum obs packages16:01
piggzrinigus:  i just need to test it doesnt break anything16:03
riniguspiggz: sure, take your time16:03
piggzrinigus: heh, i broke all the github/lab integration :)16:29
piggzjust aswell i tested such simple change!16:29
riniguspiggz: :) nice16:30
piggzrinigus: hmm, im getting "Bad Request from the github api16:51
piggzi tested the query in the explorer and it worked fine16:52
riniguspiggz: I would need ot see the change in the code to comment. could be caused by several issues...16:53
piggzrinigus: see
piggzjust popping out16:55
piggzit shouldnt actually change the strings being passed to the api, just removed the escaping in the string literals16:56
riniguspiggz: there was something with it, later in the submission of strings to api call. I am not sure whether earlier raw string haven't had escape. try to qDebug() print before your change and after16:58
rinigusI think already before it had \" in the string and it was needed somewhere else16:59
piggzyeah, just thought of them17:00
piggzi wonder why .. when i debu print it, ans paste it in explorere its fine17:00
riniguspiggz: it was a bit of a struggle to get it work. let's keep as it is and not touch that (wonderful <- irony) working solution17:02
piggzrinigus: ok .. it just fixes some warnings generated by lupdate17:02
riniguspiggz: right! but its just due to lupdate being not aware of some constructs. I would suggest to ignore it17:04
piggzrinigus: sure ... i also though it was a bit error prone becuase you acutally use the closing escape sequence in the query string eg ") ... but, i guess it works17:05
riniguspiggz: it seems to be able to get over those errors and creates TS that works17:06
piggzrinigus: yeah, only difference in output is the \ around the " in the params ... which was what i was trying to remove ...  so, may aswell just release as is17:07
riniguspiggz: question is now what would be a release version? 1.0.0 or 0.3.0? something else?17:08
piggzimok wither either17:09
piggzits just a number :)17:10
rinigus... thinking ..17:11
riniguspiggz: ok, we can go then for 0.3.0 for now. when translations catch up and we will get issues/bugfixes we can jump to 1.0 when we wish17:14
riniguspiggz: who will do the release?17:22
Thaodanrinigus bionade24: Yes you can reuse that package and update it. There is some use to have it in SailfishOS in general however I'm not sure how much it changes between kernel versions.17:28
bionade24Thaodan: I'm an absolute perf noob, just used it because callgrind was buggy. Recording and viewing the results went fine (Xperia X), but the function adresses weren't replaced with the debug symbols. Are there any additional packages I need from the repos?17:31
NicoHm, don't you just need to build with debug symbols?17:33
NicoI mostly use hotspot on my desktop, so I never bothered looking into what I need. I usually just build a RelWithDebug or debugoptimized build17:34
bionade24Nico: I did use a profile build bc I used callgrind before I'll try with debug build mode.17:35
ThaodanNico: I don't think that's needed on desktop (arch) it works without debug symbols I think.17:35
NicoWhat is a profile build? :D17:35
NicoThaodan: Well, for me some functions were missing without debug symbols17:36
ThaodanOh yeah true17:36
Thaodansome but most should be there17:36
bionade24Nico: A option in Qtcreator I haven't comapred the actual compiler flags17:36
Thaodanprofile builds are used for pgo builds to get data on how to optimize and then rebuild with that data17:36
bionade24Thaodan: What's pgo?17:37
NicoProfile guided optimization17:38
NicoWhere the optimizer knows,what branches are executed often, because you ran the binary before already and stored some profiling information17:38
piggzrinigus: ill let you  .... you know the github release process17:56
riniguspiggz: will do18:11
riniguspiggz: done, v0.3.0 released and tagged. building in chum:testing. want to write announcement to be published when OBS builds are done?18:27
piggz[m]rinigus: sure18:34
piggz[m]rinigus: do we want it in the main Announcing sailfishos:chum thread?18:41
riniguspiggz: either there or a new thread. I don't have huge preference regarding it. bit tilted towards new thread18:43
riniguspiggz: we are ready to go! OBS builds are there (chum regular and testing), all seems to be fine18:53
riniguspiggz: already points to the correct version for GUI18:54
riniguspiggz: excellent! now "just" need to add those metadata sections to SPECs19:09
piggz[m]rinigus: just added a litte info about that insecond last para19:10
riniguspiggz: :)19:11
rinigusI am going through my packages. haven't finished yet...19:12
piggzi guess i should do the same19:12
riniguspiggz: especially as Advanced Camera is the number ONE!19:20
piggz[m]rinigus: its hard to get screenshots of advanced camera, it takes control of the vol keys!19:23
riniguspiggz: but you have some at openrepos...19:24
piggz[m]rinigus: yeah, and theyre from a really old version!19:25
piggz[m]maybe there is a command line way of takng a show19:25
riniguspiggz: or compile a version without vol controls?19:27
piggz[m]rinigus: dbus-send --session --print-reply --dest=org.nemomobile.lipstick /org/nemomobile/lipstick/screenshot org.nemomobile.lipstick.saveScreenshot string:"/home/defaultuser/Pictures/Screenshots/Screenshot-$(date +%y-%m-%d19:31
riniguspiggz: handy19:32
piggz[m]shame its night, not gonna have a nice photo!19:34
riniguspiggz: you can change screenshot later by just swapping it at github repo. just use the same name and no OBS rebuilds are even needed19:37
piggz[m]yeah .. ive taken some now anyway :)19:40
piggz[m]rinigus: your crazy long build times are slowing down my checking my store entry :P20:37
riniguspiggz: I know, I know. suffering as well. haven't yet added valhalla - that one is also a great one20:38
rinigusjust hoped I could finish with those bigger ones. but probably valhalla will better wait a bit20:39
riniguspiggz: and sorry for it!20:39
bionade24rinigus: piggz[m] Why have you decided to not make a general installer like Storeman now has? The architecture selection is rather confusing, even I as dev only remembered only in the last moment, that despite my kernel being aarch64 my system's libraries are amrv7hl20:59
bionade24rinigus: piggz[m] Maybe options for the officially supported devices could be a solution in the meantime.21:01
bionade24Since /usr/lib/ --help isn't possible on SFOS I also don't know a place where users could actually detect the necessary arch easily.21:14
Thaodanbionade24: Those devices are just older ports  where aarch64 support wasn't there. Nowdays $PORT_ARCH == $ARCH21:35
bionade24Thaodan: I know, but I thiknk they're many users out there which believe uname -a returns the architecture they need or even just know my XA2 or 10 is 64bit so I guess my packages have to be too.21:37
bionade24I also find it funny that Go 64-bit binaries work with 32-bit userland but Rust 64bit binaries don't21:39
bionade24Probably because Rust still links against the libc.21:40
ThaodanAs it should IMHO. Rust and Go's staticly linking fetish is backwards.21:40
ThaodanThey have no stable abi or d_pointer style classes to emulate that.21:41
bionade24Thaodan: Yeah, not want to argue about that point. Just find it funny that Rust only does the amount of statically linking they have you bc they still don't have a stable ABI.21:42
bionade24Nico: I still get "Failed to open /tmp/, continuing without symbols" and "Failed to open [vectors], continuing without symbols"21:43
bionade24Nico: In the end all functions are still only named with their adresses, do I need to specify some option to perf report ?21:43
ThaodanYeah its weird. They have a bug open about that since 2013. I hope it gets better when GCC-Rust moves on.21:43
ThaodanAbout: perf I couldn't built it, I will try to update and see.21:44
Nicobionade24, how did you install the package?21:44
bionade24Nico: With pkcon install-local perf-<blablabla>.rpm That still resolved all dependencies and installed correctly. The OBS repo from Thaodan doesn't contain any other packages.21:47
bionade24Thaodan: Thx21:49
Nicobionade24: No, I meant the code you are debugging21:50
NicoBy default rpm strips the debug info21:50
bionade24Nico: I used the debug build options in Qtcreator. Debug symbols are definately installed.21:52
NicoHm, could be a permission issue then? No idea :D21:52
bionade24Nico: The file under /tmp doesn't exist anymore. Maybe somehow related to the fact that I have 1.5GB perf.data22:01
NicoWow, that's a lot :D22:02
bionade24Nico: Yeah, the problematic part takes > 1day, but should be possible in under 10min. I want at least some data, but I'll try with a smaller part.22:03
NicoDon't run it for a day. If it should take 10min, run it for 10 min :322:04
NicoYou can also compress the perf data22:04
bionade24Nico: I didn't run perf that long22:04
bionade24Nico: Ok my phone is weird perf crashed the kernel /o\22:04
bionade24Cool everything crashes22:46

Generated by 2.17.1 by Marius Gedminas - find it at!