Hi all,

I'd like to propose to set-up some procedure for releases.

That's not necessarily code and scripts (though in the end, the more automatized we can be, the better), but at first it could be at least just writing one wiki page saying what happens in general for a release, then we write 1 additional wiki page per release for any specifics.

This way, we can:
- organize ourselves so that work is not duplicated; - write down not-to-forget information of any manual procedure until we manage to automatize it;
- write down release-specific details like version of bundled dependencies (may be important for us, developers, to know what is used in a release), but also the list of applied patches (let's stop maintaining different patches in each build! That's a lot of boring job for you; and makes upstream life difficult if we don't know this), the build configuration (which options you used in ./configure, etc.). - not rely entirely on our benevolent maintainers (Sven and Mitch), but anyone can be a part of helping the release happen, which would mean a release better tested, but also much faster to happen.

The best example of what can go wrong is the last release: GIMP 2.8.8. We ended up apparently with major bugs on OSX Mavericks; and several fontconfig bugs on Windows, which were already fixed in the nightly builds for ages!
This kind of thing should not happen and should have been detected before release.

I took the liberty to set the pace by creating a new wiki namespace "Release". I hope that's ok and that it is not considered a bad move. And I created 2 pages in this namespace:

1/ The general information where we want to set all information about the process in general: the release procedure, all the patches to apply for every release.
http://wiki.gimp.org/index.php/Release:General_Information

2/ One page for the specific GIMP 2.8.10 release: http://wiki.gimp.org/index.php/Release:GIMP_2.8.10

Please if we have any TODO or tasks or something, just write them down there. If people wants to participate and take on a task, they could just write down their name next to an existing task and the result. This is a working page, edits should happen there!

Also I propose that before a build is officially released, they are first listed on this page for willing people to test it and an email should also be sent on the dev mailing list. If all goes well, all tasks done, and all builds are tested and look good, then we can publicly announce a release and distribute builds that have been actually tested first.

What do you think?

Jehan

On Wed, Nov 27, 2013 at 5:16 PM, Alexandre Prokoudine wrote:

27 нояб. 2013 г. 5:14 пользователь "Jehan Pagès" написал:

Also I propose that before a build is officially released, they are first listed on this page for willing people to test it and an email should also be sent on the dev mailing list. If all goes well, all tasks done, and all builds are tested and look good, then we can publicly announce a release and distribute builds that have been actually tested first.

What do you think?

I think you deserve Medal of Honour, sir :)

Hahaha. Well let's hope that others agree then. It would be good to have a better release organization.

Jehan

Alexandre
_______________________________________________ gimp-developer-list mailing list
List address: gimp-developer-list@gnome.org List membership: https://mail.gnome.org/mailman/listinfo/gimp-developer-list

Hi,

It was sent to me only. I imagine that's a small mistake, so I'll answer back to the whole list. :-)

On Fri, Nov 29, 2013 at 7:41 AM, scl wrote:

Hi Jehan,

I am ... impressed ;-)

Yes, the release procedure is something that has potential to improve and I'm happy you took the time and elaborated a procedure. I fully agree with you that we need more testing before a release. As yet I had a glance over your proposal and hope I will find more time next for a more comprehensive answer. Currently I'm fully stretched with real life tasks, getting a deeper understanding of the color topic, the UI theme and Jenkins.

Well actually the 4 main points are: 1/ testing: right now, releases are sudden, out of nowhere, and we discover release issues afterwards. Not only program bugs, but really release bugs (like the fontconfig ones on Windows). Seriously if the Windows package had been first released in beta, I would have tested these specific fontconfig issues (because I spent quite some time trying to debug these), and would have reported them immediately (so they would not have happened in the finale release). 2/ Work duplication: as you noted, many people on OSX are doing the same thing. On Windows, well there are Drawoc and Ender which have 2 different procedures too.
3/ No knowledge passing down: even though everyone knows how to compile GIMP, it looks only the package maintainers know how to make OSX/Windows installers (each with its different procedure). First it means I can't make a package for testing. But even if it were made very easy to produce installers, we could even propose users to test fixes and debug versions on Bugzilla when we are trying to debug issues. Right now we are totally dependent, and if the packagers disappear, we have to guess the procedure from the start. 4/ It looks like it is complicated for each of these individual packager. When I see for instance Simone Karin Lehmann saying that she just made a release and wouldn't do it again immediately (probably because too boring/annoying task), that is too bad. On the other hand, all together, I know we can simplify complicated tasks a lot. With scripts, automatization, and sharing work. That's what usually happens when a lot of tech people work together. Ideally it should be so easy to package a new version with a simple command, we go get a coffee and get back in 2 minutes and... tadaa!!

Eventually the procedure should find Mitch's agreement.

Of course. I just wanted to give a little push. Because just saying it would make nobody move, I feel. Maybe starting with a wiki page ready to be filled by the various packagers could be a small first start to collaborate (since people like wiki). Of course if nobody cares, well it will just stay a dead wish and I'll remove the pages in some time.

BTW: Thanks for the compliments about the benevolent maintainer. I wasn't aware that I am seen as a maintainer like Mitch (who is our real code guru here ;-), am I? Let's also not forget

Oh right. I just checked again. In the AUTHORS page, there is actually a Sven Neumann as maintainer alongside Mitch. I actually realize I made a mistake, because that's another Sven! And since I don't see him much (I think, but with people's nickname, maybe I do!), I assumed it was you and did not pay attention to the family name. :p Well good you took it well. Hopefully nobody took it bad too!

our eager packagers Jernej, Drawoc, Clayton, Simone and Partha ;-)

Of course, I have listed them in the wiki page for instance. :-) I completely understand that each individual enjoys being a maintainer — and thus the main actor! — on their own piece of code (the installers). And that's great, but that's also the reason I fear some might not like work together and share the title. Maybe? In any case, I really wish everyone could be happy by still being more efficient together. That's the goal. And I really hope we can reach it.

Anyway let's wait and see how it goes.

Jehan

Greetings,

Sven

Hi,

On Sat, Nov 30, 2013 at 11:37 PM, Simone Karin Lehmann wrote:

Am 28.11.2013 um 22:25 schrieb Jehan Pagès : Hi,

Well actually the 4 main points are: 1/ testing: right now, releases are sudden, out of nowhere, and we discover release issues afterwards.

yes, we really need a test cycle before a release goes public. Especially if there’s not only a new GIMP source, but a new OS version as well, like it is on OS X. I just discovered a new bug, which is IMO release critical. On OS X Mavericks, the pencil and brushes outline doesn’t show, so it’s almost impossible to paint, clone and brush.

Well... that's why I was proposing testing and collaborating *before* releasing, and not after. :-/

2/ Work duplication: as you noted, many people on OSX are doing the same thing. On Windows, well there are Drawoc and Ender which have 2 different procedures too.

well, I’ve tried to answer this in another thread. So let’s give it a new try.

4/ It looks like it is complicated for each of these individual packager. When I see for instance Simone Karin Lehmann saying that she just made a release and wouldn't do it again immediately (probably because too boring/annoying task), that is too bad.

It’s not about building. I wrote a couple of scripts which automates that quite well. But in the last years I’ve made a lot of OS X specific patches (don’t ask, why some of them are not upstream…. long story) and making a new release requires to adapt these patches and to test if the are still needed and if they still fix the addressed issue. Two examples: years ago on X11 it took me for ages to fix the file chooser sorting bug. Well, on Mavericks it’s back again. Second: using the Cocoa based version of gtk-mac-integration to get properly working menus and keyboard shortcuts.

Well that would still be a lot better for the users *and* for you if we could all collaborate. If these patchs on third party are really necessary to prevent major bugs, we'll appreciate having them in our source tree as well (we have a directory build/osx/ where we save build-specific data, like third-party software patches). This way, we can share these patchs will all packagers, and doing so will also save you time as we would take on us to check and adapt the patches. Could we know more about your patches, and what they fix exactly? Would you accept to contribute them to us?

All this said, the preferred politics is indeed to contribute upstream if possible. What is the reason why you did not propose your patches upstream? You can make the short version of the story if you like :-).

Further in the past I’ve tried to test that new sources fit into the „Mac standards“ and wrote patches to do so. E.g. moving the config directory to it’s proper location in ~/Library/Application Support.

Well this one is indeed useless now. :-)

New OS versions introduce new bugs. See the pencil / brush issue I mentioned above.

I saw the email and the bug report from the contributor. Have you been able to reproduce it also?

Pushing releases in such short cycles forces me to „just run my scripts“ to get a new package out and satisfy all users who start asking for the new packages. I already have a lot of requests for a SnowLeopard version. This leaves no room for testing or fixing already known issues. And that was the only thing I wanted to say when I wrote that I don’t want to redo some work. Sorry for not writing that clearly enough in the first place.

No problem. But this is exactly why it would be a lot better if you could discuss with our team OSX packager (Clayton Walker). I'm sure a collaboration into a single OSX release could save you time and allow to do better testing.
May I ask exactly what is different with your release and the ones that Clayton Walker do?

I see you add some photo editing plugins. Is that, along with the third party patches, all the difference?

Maybe you could also drop by IRC (#gimp on irc.gimp.org) and discuss with us some way to improve the situation?

Jehan

Simone Karin
_______________________________________________ gimp-developer-list mailing list
List address: gimp-developer-list@gnome.org List membership: https://mail.gnome.org/mailman/listinfo/gimp-developer-list

Hi,

On Sun, Dec 1, 2013 at 2:50 AM, Simone Karin Lehmann wrote:

Hi,

Am 30.11.2013 um 12:26 schrieb Jehan Pagès :

Well... that's why I was proposing testing and collaborating *before* releasing, and not after. :-/

yes. Sadly, now the „no-outline-issue“ is another showstopper.

It’s not about building. I wrote a couple of scripts which automates that quite well. But in the last years I’ve made a lot of OS X specific patches (don’t ask, why some of them are not upstream…. long story) and making a new release requires to adapt these patches and to test if the are still needed and if they still fix the addressed issue. Two examples: years ago on X11 it took me for ages to fix the file chooser sorting bug. Well, on Mavericks it’s back again. Second: using the Cocoa based version of gtk-mac-integration to get properly working menus and keyboard shortcuts.

Well that would still be a lot better for the users *and* for you if we could all collaborate. If these patchs on third party are really necessary to prevent major bugs, we'll appreciate having them in our source tree as well (we have a directory build/osx/ where we save build-specific data, like third-party software patches).

yes, I’ll share them. But IMO this needs to get committed about what build system we use on OS X. Clayton uses jhbuild. I use a slightly hacked version of MacPorts and some bash scripts to ease the process of building on Mavericks down to SnowLeopard and even cross compile to 32bit and it helps me manage about 100 packages needed to build my bundles. As far as I’m concerned, I’d like to stay with that and not to switch to something else I’m not used to and from what I don’t know if it fits my needs and gives me all functionality I already have.

Well, this is up to you to decide. I know everyone of us, developers, think we are right, at least at first. And maybe you even really are (= maybe your building solution is nicer than Clayton's, I just have no idea). But in the end, being right does not matter compared to the benefits of collaboration. GIMP would be nothing compared to what it is now if it had stuck to be a single-man project. Stubbornness is usually a good thing... up to one point. :-) But yes in the end, you are to decide what you want to do.

This way, we
can share these patchs will all packagers, and doing so will also save you time as we would take on us to check and adapt the patches. Could we know more about your patches, and what they fix exactly?

e.g.
glib, gtk2, cairo
using Coca instead of Carbon, fixes paths to fit into Mac standards, etc. gtk-mac-integration:
use Cocoa, fix some issues, don’t hide the delegate and notification protocols to enable easier app development on the application side. gimp:
Cocao, working help system with reduced config options, using some system provided libraries instead of build them from source, Mac shortcuts, lightly different behavior of lcms to recognize more icc profiles, working dock menus :-), hide / unhide Gimp, and a „right-out-of-hell-and-never-will-be-included-patch“ about the save / export issue ;-)

Here’s the link to the current epository (not totally complete, I’ll update this if I find some time…. and I know, some code looks ugly …)

https://sourceforge.net/p/gimponosx/code/HEAD/tree/

Thanks. I'll see what Clayton thinks about these.

Would you accept to contribute them to us?

if we could negotiate an what to use …. :-)

Well on our side, we have not much to give, or "negotiate". We just want to collaborate with as much packagers as possible, so that they become actual upstream contributors and save everyone's time, but also give a much better user experience to all users.

All this said, the preferred politics is indeed to contribute upstream if possible. What is the reason why you did not propose your patches upstream? You can make the short version of the story if you like :-).

hhhmm, what should I say, I don’t want to revive these zombies, but I’ll try a short version. (you’ve been warned :-) )

In the „old days“ of the Mac packaging community most things were fine. But then patches or plugins got rejected with a simple „No, we don’t include this, because we are the official packagers“. Other patches were silently taken and rewritten without asking why I did it in this specific way. No discussion about why I did things or how to improve things, all of a sudden, everybody only wanted packages. No one was interested in going deeper. Although I asked for help to fix a long standing issue (using the help system, install the user manual locally, get rid of the „GIO/GFVS“ error. (BTW, all of this is now fixed, it took me years…)) … nothing happened.

Well I wish you could also see the other side of the story. Now I don't know your exact cases of course, but I am a Free Software contributor too. And I got all sort of disappointment with it too. I got a lot of rejected patches in my life and the developers would not seem to care about my explanations. Well maybe they simply had another vision of the software as I. And as I said before, the problem is not to be right or wrong. Who am I to say they are? I even got patches totally ignored, even though I repeatedly came back and abandoned in the end. Very frustrating too. I also got rewritten patches in ways I disagreed, or even patches taken as is, but with authorship changed to a developer's name! Well that's all too bad, but that's life. Developers/maintainers are all human, just like you. Some have not enough time and forget your patches or misread them, some are not that skilled, some are actual asses (doing FOSS does not prevent you from being an ass), some just forgot to check a patch because they are not well organized, some simply disagree and even though they would like to please anyone, you can't always do so, and so on. Also all can do mistakes, even the best ones! That's important to note.

Basically it's hard to contribute to FOSS, because it's still human relationship, and it requires some patience on everyone's side. But the rewards is very good. There is still so much you can do alone. And in the end, working upstream instead of maintaining what are basically mini-forks of that many programs is so much powerful: you touch all users (and not only a small user-base), you are ensured your changes will survive any rewriting, you don't have to always maintain them… In other words, you can carry on, proceed to other issues at hand and forget about these past issues. That's so much better than constantly having to maintain the same boring past issues, which is extremely limiting and like being glued to a status-quo. So I really hope you will reconsider contributing upstream, not only for GIMP but for the many projects you seem to have patches for. Once again, I think none of the widly used FOSS programs would be even half of what they currently are if all contributors had stuck to themselves and made forks.

I got the impression that, with a few exceptions, nobody wants to contribute to the Mac version.

I think the main point is that there are less Mac developers in Free Software historically. They are rather using Free OSes (Linux, BSD, etc.). So it's not that they don't want it, it's just they don't use, so we just wait for a contributor actually on this OS to step in. This said, it changes, and I see more FOSS developers with a Mac.

And then came this discussion about a „native“ build. Everybody thought that a native version will automagically solve all problems they have. But „native“ is IMO much more than simply dropping X11 and have a menu bar at the top. It’s about using OS X functionality like ColorSync, the print system, system provided libraries, using Cocoa not Carbon.

I don't think many people would disagree with you. It is obvious that integration in any OS is more than just doing a copy-paste from another platform. I think there is enough OSX-specific code in GIMP to show that we believe it too.
But we still need to start somewhere. And this start is at the common code, that can be further specialized. Also we want to do things well, which still means have as most common code as possible, and push down specific code to layers down (GTK+, glib, etc.). Doing things well takes time but ensures less bad surprises in the future. In any case we welcome any patches to go further in integration!

Again, I got the impression that there are only very few people interested in contributing to the Mac version. Most people only wanted to build a package and to have the menu bar at the top. So I decided to do it on my own. Last zombie: for a short time the link to my site was dropped in favour of a „native" build with the menu bar on top. And although GIMP now being "officially native“, well, there are only few things from other OS X developers and packagers to port more code to native OS X functionality. But now let the zombies rest in peace and give the OS X version a new try.

New OS versions introduce new bugs. See the pencil / brush issue I mentioned above.

Well bugs, we won't ever totally get rid of them. This is nearly by definition! :-) There are bugs on GIMP, whatever the platform. :-P That's why we welcome contributors even more.

I saw the email and the bug report from the contributor. Have you been able to reproduce it also?

sadly, yes.

No problem. But this is exactly why it would be a lot better if you could discuss with our team OSX packager (Clayton Walker). I'm sure a collaboration into a single OSX release could save you time and allow to do better testing.
May I ask exactly what is different with your release and the ones that Clayton Walker do?

mainly the above mentioned patches, a lot of plugins, my packages are signed with an Apple Developer ID.

Maybe you could also drop by IRC (#gimp on irc.gimp.org) and discuss with us some way to improve the situation?

puuh, I don’t like IRC because of my bad english (I’m just to slow to keep up…) and it’s mainly about not having enough time. Sorry. I try to do my best, but now I need to do some other work (keeping the house clean and prepare for a business meeting and trip next week). Again, I’ll try to collaborate as much as I can.

I don't really like IRC either, mostly because I don't like much chatting. Or rather the opposite: I don't like to have to "wait" there and nothing happening (I prefer people to send me a message, IM or email, when they need it). But that's how they do in GIMP. So I am not *always* in the IRC channel, but I go sometimes when I think about it. Often indeed nothing happens. But sometimes, well I can get some shit done because I was there. So I feel better. Again sometimes collaboration means you have to do compromises because this is a 2-way thing. :-)

Jehan

Regards
Simone Karin

—
stay hungry, stay foolish

_______________________________________________ gimp-developer-list mailing list
List address: gimp-developer-list@gnome.org List membership: https://mail.gnome.org/mailman/listinfo/gimp-developer-list

Hi,

On Sun, Dec 1, 2013 at 12:44 PM, Jehan Pagès wrote:

On Sun, Dec 1, 2013 at 2:50 AM, Simone Karin Lehmann wrote:

Here’s the link to the current epository (not totally complete, I’ll update this if I find some time…. and I know, some code looks ugly …)

https://sourceforge.net/p/gimponosx/code/HEAD/tree/

Thanks. I'll see what Clayton thinks about these.

We had a quick look with Mitch on the patches for GIMP specifically. Many of them are obsolete now. For instance we already switched to the Cocoa platform since 2.8.10 (commit
e56344294c90e1ba97de5c134b50c4c522f0808f which includes code from you already!).

And some like the save/export, we obviously won't integrate, as you can guess. But the "color display change" for instance seems promising. :-) If you have other contributions you would like to see upstream, I would really suggest discussing them with us and providing us patches. This would really improve everyone's process! Thanks anyway.

Jehan

Hi,

On Tue, Dec 10, 2013 at 5:00 AM, Sam Gleske wrote:

On Mon, Dec 2, 2013 at 11:09 AM, Sam Gleske wrote:

What are the Windows packagers using for the build? If they're using a proprietary package system I suggest moving to a more open one such as NSIS and customizing that. Windows builds can easily be automated using NSIS. Another advantage of using NSIS is checking the scriptable installer code into SCM.

Does this mean that nobody on this list knows what the release process is on Windows for building installers?

As Michael answered too, of course we have people in charge of the Windows release procedure. Simply I am hoping we can improve it by merging efforts, since several people are also known to do their own quality release outside us. And our procedure is far from perfect. I wish it were more systematic and without a fault.

But we do have an existing procedure since we do have releases and code for it in our repo. And we are slowly improving it (for instance you may have noticed the recent improvement: now all releases go in the ftp, instead of using a third party provider).

Jehan

_______________________________________________ gimp-developer-list mailing list
List address: gimp-developer-list@gnome.org List membership: https://mail.gnome.org/mailman/listinfo/gimp-developer-list