Monday, 2020-08-31

T42<linusdan> Hi, guys! I have a question: Is it necessary to add something to make non neon devices compatible with Sailfish apps or is android hal enough?01:20
r0kk3rznon neon devices like what02:06
r0kk3rzsailfish rootfs only really comes in two flavours02:07
rinigusmal: I'll wait for your approval04:13
riniguspiggz: several goals here: help Nemo, check if latest obs is ok (should reduce maintenance), and get maps apps into obs with regular Linux04:24
rinigusNot much pushing Jolla in this respect :)04:25
r0kk3rzi got the impression that they've already decided to kill our obs04:26
rinigusSimilar impression here. Question is then what will be instead, but that seems to be unknown. Maybe we can get help with setting up another OBS or reconfigure current one04:55 admins: help! can you please increase the quota for number of projects to me on the server? as I fix trivial build issues, I may need way more projects than is available now05:45
rinigusmal: that exe bit is needed also to fill properly provides for rpm. without exe bit set, library is not listed in "provides" of the package on fedora08:25
T42<adampigg> linux - Why are .so files executable? - Unix & Linux Stack Exchange08:33
spiiroinI blame rpm. IIRC etc, rpm build uses "locate +x bit files" as substitute for proper "find elf binaries" -> stripping debug symbols etc fails if libraries do not have x bit set08:37
rinigus@adampigg, mal:
spiiroinyou can arrange it so that final packaged libs do not have x bit, but ... whether it is worthwhile08:38
rinigusspiiroin: according to the link above, it is default for gcc to set it executable08:38
spiiroinrinigus: well, it makes no sense unless it is very special library - you are not supposed to execute them08:39
rinigusspiiroin: as for blaming rpm, you are probably right. although it does seem to be default on rpm distros to have it set. and sfos is using rpm08:39
r0kk3rzmoo spiiroin, is it important to build stuff with sb2 for sailfish, or would any cross compiler do?08:39
rinigusas for whether it makes sense, I agree with you08:39
spiiroinr0kk3rz: I guess it depends, there are some projects that use sailfish sw + completely different build envs and there is no harm in it as long as both can peacefully co-exist08:45
spiiroinlike, if there are prs related to supporting other tools, sb2/obs/whatnot should keep on working08:46
r0kk3rzspiiroin: we were thinking about using OBS but with its kvm based cross compiler instead08:46
r0kk3rzand I wasnt sure whether that was a terrible idea or not, since theres presumably some benefit to using sb208:49
r0kk3rzor if sb2 is just used for hysterical reasons08:49
spiiroinr0kk3rz: personally I use only sb2. local builds. one component at time. then occasionally fix something if builds in obs fails after tagging...08:50
spiiroinbut that is probably just because it is rather seldom that I need to touch more than one component simultaneously08:51
r0kk3rzyeah, but thats only because you guys have your own OBS chugging away in the background :P08:53
rinigusspiiroin & mal: (as it helps to have 2 discussions at the same time) so, any decision regarding +x set for libs? as we already have it for the most of the libs anyway08:56
spiiroinrinigus: IMO based on what I think I know: not having +x in rpm build system is fighting windmills08:58
rinigusit does feel so08:59
spiiroinI might have done something somewhere to avoid it, but I would not require anybody else do it as it does require going through otherwise unnecessary hoops and loops when most if not all packages still have the +x09:00
rinigusthanks! I have submitted 2 MRs last night: one for setting mer-dev to noarch (otherwise fedora looks for exe bits and complains about debug symbols) and one for libglibutil with +x09:02
rinigusmal wanted to check something as well before giving the green light, as far as I know09:03
piggzrinigus: top effort, will try to get on board soon to help out09:03
rinigusif all is OK, I can start submitting MRs as I encounter them09:03
riniguspiggz: thanks!09:04
rinigusspiiroin: who to ping to get quotas lifted?09:04
spiiroinrinigus: lbt ^09:05
malrinigus: the thing I meant was just about bug references in the commit messages09:16
malmeaning what kind of bug ref to use in those09:16
rinigusmal: understood. I presume you want me to have the same ref in all of these trivial commits?09:18
rinigusor do you use it in the merge commit?09:18
malrinigus: use JB#51013 as bug reference in the executable bit commits10:48
rinigusmal: will do. I will then change the commit message for the already submitted MR10:50
rinigus... tonight10:50
malrinigus: that was merged already10:51
rinigusmal: for some reason all messages from gitlab end up in spam :(10:53
riniguswill set the filter10:54
malrinigus: this won't work for us, maybe try to add --disable-static to configure line in the spec16:44
maland then remove the new additions to files (sfos build removes the static files anyway)16:44
riniguswill do. I will close that PR then and make a new one16:44
rinigusmal ^16:45
malyou can fix the PR16:45
malunless you for some reason want to use a different branch16:45
rinigusOK, will fix PR then. no problem with the branch16:46
rinigusjust hit gitlab limit again, but let me fix the PRs I have16:46
rinigusmal: disabling static and dropping .a was sufficient. is it OK now?17:07
*** Oksana_ is now known as Oksana17:25
malrinigus: it still generates .la files?17:47
rinigusmal: it should, otherwise it would fail17:49
riniguslet me download rpm17:49
malrinigus: I think that will still be a problem because sfos build won't have those17:49
malrinigus: did you make test builds on both OBSes?17:50
rinigusmal: I was wondering about it. I can add %if 0{?fedora} in this case17:51
rinigusmal: no, I didn't test on SFOS one as I had issues with compilation of nemo core packages on it (tried on one of them). could be related to heavy project config used there17:51
rinigusmaybe SDK would be more adequate in this case17:52
malrinigus: in many of our spec files we have or some have "rm -f %{buildroot}/%{_libdir}/*.la"17:53
malmaybe not many but some17:53
rinigusmal: that should do, I will do it that way. it sounds like these files are not needed anymore17:53
malrinigus: so if there are no subfolder then "rm -f %{buildroot}/%{_libdir}/*.la" is probably simpler17:54
rinigusmal: agree, let's keep it simple17:54
malfind is needed for more complex builds17:55
malrinigus: platform sdk build is also a good way to test17:55
rinigusmal: submitted cleaned up commit. sorry for not testing it before with SFOS, should have done so18:36
malrinigus: looking better now19:04
malrinigus: also fix the PR title19:05
rinigusmal: done19:11
pketorinigus: I increased your project limit on (cc: lbt)23:17

Generated by 2.17.1 by Marius Gedminas - find it at!