Hi,

On 10/03/16 11:27, Peter Uhn��k wrote:

Sorry I don't explain myself properly. I was meaning a window with this message instead of a error trace, which could be intimidating. Something with the same message and buttons like "Ok", "Launch debugger". In the workshops there are some Gnu/Linux newbies and the hidding the error trace for a while seems to be helpful to them.

This should be relatively simple to do:

1) create new "PreDebugger" that's user-friendly (currently we have (GT)SpecPreDebugWindow that is started from (GT)SpecDebugger>>openWithNotification:)
2) let Debuggers ask for PreDebugWindow class through Smalltalk tools
3) install your UserFirendlyPreDubugWindow to Smalltalk tools

I think this would be a very useful addition for many people. (I wanted to make this some time ago but haven't got around to do it yet.)


Thanks for the ideas.

No, but because we could build a wrapper for particular packages that installs Nix, and then others (32 bits variants of libcairo or sqlite, pandoc, etc) without caring about a particular Unix variant (being them Mac or hundreds Gnu/Linux distros based on dozen of "bases"), making the experience smother for the newbie, providing he/she has root privileges. People will not see nix anywhere, not even for installing

Mac is self-contained, so the main issue is linux.
If the user has a root privilege, then maybe we should push for pharo directly in the package manager repository?

What I would like is to push away the users from the environment and computing experience as little as possible, to make the experience work for them. That mean minimizing go to the web, go to command shell and so on for things like installing the stuff that make pharo/roassal work.

And sure, there are hundreds of distros, but you will cover majority of the market with��� maybe 5 distros? People that can't handle obscure distros don't install them.

But to get some motion:
1) let's compile a list of majority distros
2) let's see how hard it is to build packages for them (imho this could be automated via CI), and we already have ubuntu package (or maybe nix can be pushed to all distros?)

Nix can be used on all Linux distros, Solaries, Mac and major close to Unix operative systems, so it is already dealing with the diversity without us reinventing that an using it where complex installations and dependencies are needed seems a sensible option.

3) maybe consider all-in-one bundle for non-root targets (this would work for cairo, but not for 64b systems in general, because the dependency list there is monstrous)

That bundle would be really helpful. Disk space is cheaper that user time/frustration. I'm starting some small installer for self-contained non-root apps that I need in grafoscopio (SQLite, fossil and pandoc), so when they're needed but not present, the systems deals with the installation in a smooth way, and the person keeps focused on the task (or can take a break while the installation of bigger stuff :-) ).

��

So no need to be sarcastic here Peter.
��
Sorry, I got carried away. I've got my fair share of dependency hell and introducing another package manager always makes me shiver.


Don't worry. In my case, Nix has been a savior when a dependency hell can not be solved my my native package manager, so I see the things in the opposite way.

Cheers,

Offray