Apt overlay tech landing in neon/unstable momentarily. Going to move to user editions if unstable testing reveals no major snags.

Approved on telegram by Jon. Turns out the problems were something gone wrong with patching. Overlay landing in unstable for preliminary testing

drop argv duping. it's not necessary

Ah yes. The problem with kde4 is actually that even if releasme downloaded the assets correctly, they'd not get used. In kde4 times the release tools were supposed to inject the relevant cmake logic to build and install po/ and docs/. We've moved away from this because it's horrible to maintain and it divides the git-source from the tarball-source in an unnecessary fashion. But its horribleness was the reason modern releaseme intentionally does not support kde4 l10n because it'd entail a substantial maintenance cost for next to zero gain. i.e. supporting kde4 l10n does require also supporting correctly constructing cmakelists for kde4 and all that, and supporting that is absolutely terrible.

fix style for consistency with rest of file

• defaultFilePath is now static in the cpp only, the public ctor defaults to QString() for the filePath, which gets checked in the private ctor and diverted to defaultFilePath if necessary
• Private ctor now copies filePath as we may need to change it to defaultFilePath
• = and \# parsing revisitede
  • new helper splitEntry mids the line at first = character only preventing = characters in pretty strings from tripping up the parsing
  • \# characters that are not the first character in the line are now not explicitly causing a line skip as far as entry parsing is concerned. I think this is actually a more truthful implementation of the format specification. since the spec doesn't even mention trailing \# we can do whatever with them and not skipping entire lines outright seems like the nicer solution here
• api docs updated accordingly. now clearly states that trailing comments result in undefined behavior
• test and fixture updated to make sure = and \# work in pretty string and that completely invalid lines are skipped
• parsing skips now produce categorized warnings

The only relevant change in the history not from me is from @davidedmundson it seems. Dave, are you ok with changing the license from GPL to LGPL?

Nevermind found it. jenkins-jobs/nci/project.rb has special handling which mangles dependees for various criteria. Among them if it is a fork. I am not sure what this task is on about though @jriddell. It only needs the fork removed from the array?

Implemented since February in c57ff198713957b3635ae7047af3977fae284259

Oh, on a related note, I actually do wonder if we should symlink start-here-kde in breeze to start-here-kde-plasma and use that instead to differentiate? As per the spec on !breeze icon themes start-here-kde-plasma would fall back to start-here-kde if the plasma icon doesn't exist IIRC. So it's just as portable but less ambiguous perhaps.

before

New unstable ISO built using the changed livecd-rootfs and has the preinstalled pool correctly set up and packages correctly available. I'm considering this fixed.

Do we want this released pronto?

landed and seems to work fine on pristine user edition installation

By updates stable builds you mean trigger a build?

This looks to have been fixed.

I don't understand this. I also can't find the code in question or any trace of it in the git history :/

I have the change ready. waiting for next week so I don't have to rush this through on a friday and then end up with broken isos

• doc wording
• disable cctor. no real use case for copy at this point

LGTM technically. Also looks crazy scary invasive though from a packaging POV :O

Still fairly messy and very verbose but working.

Ha, this is also much nicer looking. I love it! :)

I am taking silence as indifference.

Software that hasn't been ported by now IMO deserves no further fixes anyway. At this point it's more a matter of anyone finding the time and energy to actually rip Qt4 out and moving to ECM :/

Q_FALLTHROUGH is too new. This thing is still qt 4 compatible

Fair enough. LGTM :)

Something is broken with the archives which prevents ISOs building on account of not finding neon-live. I am not clear why, @jriddell did change some stuff though and at a glance that seems at fault since the thing that didn't change (user edition) did build fine the other day. So that blocks ISO rebuilds for stable (and also breaks unstable now).

I should add for the record that I'd actually find stricter versioning better, but arg handling in kcrash is already a bit tricky because of signal-safety restrictions and adding versioning complexity on top would probably do nothing for how easy the code base is to approach. On top of that, even with minimal options (correctly) set, drkonqi will and should be able to help the user file a bug report. In theory we need as little as the pid to do something of use.