Monday, February 10, 2020

Working with different remotes in git

One of the things that is typical when working with gitlab/github is work with different git remotes.

This is sometimes because you don't have commit access to the original repository so you fork it into your own repository and work over there, but you still want to have the original repository around so you can rebase your changes over it.

In this blog we will see how to do that with the okular repository.

First off, we start by cloning the original repository

Since we don't know the URL by memory, we go to and press the Clone button to get a hint, if we have commit access we can use both urls, otherwise we have to use the https one, for the sake of this let's assume we do not have commit access.

$ git clone
$ cd okular

Ok, at this point we have clone the upstream Okular repository, we can see we only have one remote, called origin

$ git remote -v
origin (fetch)
origin (push)

Now we want to do some fixes, since we can't commit into the main repository, we need to fork, for that we press the fork button in Once done we end up in the fork of Okular under our name, e.g.

Now what we want is to add our remote to the existing one, so we press the blue button (here we use the git@ one since we can always commit to our fork)

$ git remote add aacid_fork
$ git remote -v
aacid_fork (fetch)
aacid_fork (push)
origin (fetch)
origin (push)

So now we have a remote called aacid_fork that points to url fork, aacid_fork is the name i chose because it's easy to remember, but we could have used any name we wanted there.

Now there's several things one may want to do

Do changes in master and push them to your fork

This is really not the recommended way but since it's what i do and it'll explain how to push from one branch name to another i'll explain it.

After doing the changes and doing the typical git commit now we have to push the changes to our aacid_fork, so we do

git push aacid_fork master:adding_since_to_function

What this does is push the local branch master to the branch named adding_since_to_function of the aacid_fork remote

Create a branch and then push that to your fork

This is what you should be doing, so what you should do is

git branch adding_since_to_function

and then change to work on that branch

git switch adding_since_to_function

After doing the changes and doing the typical git commit now we have to push the changes to our aacid_fork, so we do

git push aacid_fork adding_since_to_function

What this does is push the local branch adding_since_to_function to a branch with the same name of the aacid_fork

Get a branch from someone else's remote and push to it

Sometimes some people say "hey let's work on my branch together", so you need to push not to origin, not to your fork but to someone else's fork.

Let's say you want to work on joliveira's gsoc2019_numberFormat branch, so you would need to add his remote

$ git remote add joliveira_fork
$ git remote -v
aacid_fork (fetch)
aacid_fork (push)
joliveira_fork (fetch)
joliveira_fork (push)
origin (fetch)
origin (push)

Then we would need to tell git, hey listen, please go and read the branches that remote i just added has

git fetch joliveira_fork

Next we have to tell git to actually give us the gsoc2019_numberFormat branch, there's lots of ways to do that, one that works is

git checkout --track joliveira_fork/gsoc2019_numberFormat

This will create a local gsoc2019_numberFormat from the contents of the remote branch joliveira_fork/gsoc2019_numberFormat and that also "tracks" it, that means that if someone else does changes to it and you do git pull --rebase while on your local gsoc2019_numberFormat, you'll get them.

After doing the changes and doing the typical git commit now we have to push the changes to the joliveira_fork, so we do

git push joliveira_fork gsoc2019_numberFormat

What you don't want to do

Don't push to the master branch of your remote, it's weird, some people do, but it's note really recommended.

Things to remember

A git remote is another repository, it just so happens that it has "similar" code, but it's a fork, so you can push to it, checkout branches from it, etc.

Every time you want to get changes from a remote, remember to git fetch remote_name otherwise you're still on the "old" snapshot form your last fetch.

When git pushing the syntax is git push remote_name local_branch_name:remote_branch_name

Bonus track: Using git mr

As shown in my previous blog post you can use git mr to easy download the code of a mr. Let's use as example Okular's MR #20

You can simply do git mr 20 and it will create a local branch named mr/20 with the contents of that MR. Unfortunately, if you want to commit changes to it, you still need to use the original remote and branch name name, so if you do some changes, after the git commit you should do

git push joliveira_fork mr/20:gsoc2019_percentFormat

Sunday, February 09, 2020

20.04 releases schedule finalized

It is available at the usual place

Dependency freeze is in ~five weeks (March12) and Feature Freeze a week after that, make sure you start finishing your stuff!

Thursday, February 06, 2020

New help porting to Go >= 1.12

At KDE we have a GitHub mirror. One of the problems about having a mirror is that people routinely try to propose Pull Requests over there, but no one is watching, so they would go stale, which is not good for anyone.

What no one? Actually no, we have kdeclose, a bot that will go over all Pull Requests and gracefully close them suggesting people to move the patch over to KDE infrastructure where we are watching.

The problem is that I'm running that code on Google AppEngine and they are cutting support for the old Go version that it's using, so I would need someone help me port the code to a newer Go version.

Anyone can help me?

P.S: No, I'm not the original author of the code, it's a fork of something else, but that has not been updated either.

Update: This is now done, thanks to Daniele (see first comment). Mega fast, thanks community!

Sunday, February 02, 2020

QCA cleanup spree

The last few weeks I've done quite a bit of QCA cleanup.

Bit of a summary:
* Moved to KDE's gitlab and enable clazy and clang-tidy continuous integration checks
* Fixed lots of crashes when copying some of the classes, it's not a normal use case, but it's supported by the API so make it work :)
* Fixed lots of crashes because we were assuming some of the backend libraries had more features than they actually do (e.g. we thought botan would support always support a given crypto algorithm, but some versions don't, now we check if the algorithm it's supported before saying it is)
* Made all the tests succeed :)
* Dropped Qt4 support
* Use override, nullptr (Laurent), various of the "sanity" QT_* defines, etc.
* botan backend now requires botan2
* Fixed most of the compile warnings

What I probably also did is maybe break the OSX and Windows builds, so if you're using QCA there you should start testing it and propose Merge Requests.

Note: My original idea was to actually kill QCA because i started looking at it and lot of the code looked a bit fishy, and no one wants crypto fishy code, but then i realized we use it in too many places in KDE and i'd rather have "fishy crypto code" in one place than in lots of different places, at least this way it's easier to eventually fix it.

Tuesday, January 28, 2020

The Qt Company is stopping Qt LTS releases. We (KDE) are going to be fine :)

Obvious disclaimer, this is my opinion, not KDE's, not my employer's, not my parents', only mine ;)

Big news today is that Qt Long-term-supported (LTS) releases and the offline installer will become available to commercial licensees only.

Ignoring upcoming switch to Qt6 scenario for now, how bad is that for us?

Let's look at some numbers from our friends at repology.

At this point we have 2 Qt LTS going on, Qt 5.9 (5.9.9 since December) and Qt 5.12 (5.12.6 since November).

How many distros ship Qt 5.9.9? 0. (there's macports and slackbuilds but none of those seem to provide Plasma packages, so I'm ignoring them)

How many distros ship Qt 5.12.6? 5, Adélie Linux, Fedora 30, Mageia 7, OpenSuse Leap 15.2, PCLinux OS (ALT Linux and GNU Guix also do but they don't seem to ship Plasma). Those are some bigger names (I'd say specially Fedora and OpenSuse).

On the other hand Fedora 28 and 29 ship some 5.12.x version but have not updated to 5.12.6, Opensuse Leap 15.1 has a similar issue, it's stuck on 5.9.7 and did not update to 5.9.9 and so is Mageia 6 which is stuck on Qt 5.9.4

Ubuntu 19.04, 19.08 and 20.04 all ship some version of Qt 5.12 (LTS) but not the lastest version.

On the other a few of other "big" distros don't ship Qt LTS, Arch and Gentoo ship 5.14, our not-distro-distro Neon is on 5.13 and so is flatpak.

As I see it, the numbers say that while it's true that some distros are shipping the latest LTS release, it's not all of them by far, and it looks more like an opportunistic use, the LTS branch is followed for a while in the last release of the distro, but the previous ones get abandoned at some point, so the LTS doesn't really seem to be used to its fully potential.

What would happen if there was no Qt LTS?

Hard to say, but I think some of the "newer" distros would actually be shipping Qt 5.13 or 5.14, and in my book that's a good thing, moving users forward is always good.

The "already released" distros is different story, since they would obviously not be updating from Qt 5.9 to 5.14, but as we've seen it seems that most of the times they don't really follow the Qt LTS releases to its full extent either.

So all in all, I'm going to say not having Qt LTS releases is not that bad for KDE, we've lived for that for a long time (remember there has only been 4 Qt LTS, 4.8, 5.6, 5.9 and 5.12) so we'll do mostly fine.

But What about Qt 5.15 and Qt 6 you ask!

Yes, this may actually be a problem, if all goes to plan Qt 5.15 will be released in May and Qt 6.0 in November, that means we will likely get up to Qt 5.15.2 or 5.15.3 and then that's it, we're moving to Qt 6.0

Obviously KDE will have to move to Qt 6 at some point, but that's going to take a while (as example Plasma 5 was released when Qt was at 5.3) so for let's say that for a year or two we will still be using Qt 5.15 without any bugfix releases.

That can be OK if Qt 5.15 ends being a good release or a problem if it's a bit buggy. If it's buggy, well then we'll have to figure out what to do, and it'll probably involve some kind of fork somewhere, be it by KDE (qt already had that for a while in ancient history with qt-copy) or by some other trusted source, but let's hope it doesn't get to that, since it would mean that there's two set of people fixing bugs in Qt 5.15, The Qt Company engineers and the rest of the world, and doing the same work twice is not smart.

Sunday, January 26, 2020

git mr: easily downloading gitlab merge requests

With KDE [slowly] moving to gitlab you may probably find yourselves reviewing more gitlab based patches.

In my opinion the web UI in gitlab is miles better, the fact that it has a "merge this thing" button makes it a game changer.

Now since we are coming from phabricator you have probably used the arc patch DXXX command to download and locally test a patch.

The gitlab web UI has a a link named "You can merge this merge request manually using the command line" that if pressed tells you to

git fetch "" "patch-from-kde-bug-415012"
git checkout -b "sander/okular-patch-from-kde-bug-415012" FETCH_HEAD

if you want to locally test

That is *horrible*

Enter git mr a very simple script that makes it so that you only have to type

git mr 80

P.S: If you're an archlinux user you can get it from AUR

P.P.S: Unfortunately it does not support pushing, so if you want to push to that mr you'll have to do some work.

Thursday, January 16, 2020

KPatience added to flathub. Which app should be next?

This week we added KPatience to flathub.

That makes for a quite a few applications from KDE already in flathub

Which one do you think we should add next?