Porting code to Sailfish

Applications
1

I’d like to know if it makes sense to write a set of python scripts to port open source code to compile with Sailfish.

For instance, I’d like to get a XMPP client that works natively on Sailfish but looking at the QXmpp project, this is now designed for Qt6 while SFOS is blocked with Qt 5.6.

With AI this is quite easy to get simple scripts that modify files in a directory.
Sometimes, you need to emulate classes that do not exist in Qt 5.6 (for instance QStringView). A script that adds an emulation class, replaces all occurrences to QStringView with a QEmuStringView class and adds at the top of the source file a #include will be great.

This way it would be easier to get more native applications for Sailfish.

This could may be also applied to QML based code.

Do you see this as a waste of time? Do you think this is a dead-end? Does something already exist? Or do you see an alternative?

1 Like
2

If you have top ask, you know the answer.

3

i would say just do it, you will see the results.

1 Like
4

…and can adapt along the way to maybe get something useful. But asking others upfront what you should do does not exactly smell of success.

5

Dont listen to the naj’ers. SFOS is fun to develop for, but its true that the rewards are not necessarily gratifying, definitely not financially. Perhaps it depends for the reason. If the app you want to port for example is a deal breaker for you to use SFOS, id suggest give it a try! You will learn alot. The forum is full of people who are willing to help; its not an easy ride. So remember to ask for help, and you might succeed! :slight_smile:

3 Likes
6

Read a headline recently that said
“Coders use time saved by using AI to fix AI errors”
Let that be my 2ct.

edit: in essence this means that everybody who “uses” AI is a willing beta tester for big corporations who are trying desperately to get out of the red numbers with a product that simply isn’t there yet - and won’t be for a long while. Not to speak of all the implications for privacy, intellectual property etc.

6 Likes
7

AI generated code is a maintenance burden - to say the least.
Often it is not maintainable at all.

Especially in a small ecosystemlike ours with very few developers that is important as no one will be able to pick up maintenance after it inevitably breaks and the original ‘developer’ has left.

4 Likes
8

The idea is that if you need a change, you dont make it on the code. You just rewrite the whole thing with AI..but it doesnt work.

9

it doesn’t work

Doesn’t it? it would be like libhybris but for qt6->qt5, if you can make it work we wouldn’t need all the optqt5.15/optqt6 efforts, let the ai loose, if it works we’d be golden and all the complaints about old qt would stop, worth a try

10

One can add as many layers as one wishes; however, the rule of “less is more” is still something that applies… :wink:
Also, libhybris is like an automatic clock, built by a clockmaster with years of experience. Whereas a wrapper qt5→qt6 concocted on the fly by some LLM model (tuned on C/C++ and javascript) is like an old seiko digital clock, but that shows the time it wants….and will probably end up written in C#.. /jokes:off

I give you that, it might help as it has a clear input and a clear output; however, at that point one could ask, why have qt as a framework at all? Why not let the AI figure out the best framework (and implement it)? The answer is simple : its not there yet :slight_smile:

1 Like
11

No, the complaints wouldn’t stop. Who do you think is doing security fixes for Qt 5.6? No one is.

SailfishOS is being presented as privacy focused and sure, it’s private by design, but all the privacy in the world is useless if a script kiddie can abuse a vulnerability that’s been fixed a decade ago.

Sure, having new features from newer Qt would be really nice (and I wish for it every time I have to use the ancient JavaScript flavour in QML), but the real reason we need it is security.

5 Likes