@Chris

Only had a brief look, but if there is a crash during exit, is possible pending IO checked and aborted? I notice SendIO() being used.

@broadblues

Top command on Linux lists itself too.

Joerg wrote a "top" command ~10 years ago. Today I found it from my harddrive. It lists also task switches per second and apparently the time spent inside Forbid(). This version also lists itself and support dynamically started new tasks without issues.

Hieronymus unfortunately crashes on my Sam440.

@broadblues

It's supposed to be a top clone. Thus, it is obviously very useful to see at least the "top" task's name listed in the view, when there are more than one "Shell Process".

It's still one process that wants to have its unique name (see above), and the same process that wants to list the names of all tasks.

I will have to investigate exclusive startup. Due to how this clone works at the moment, it would be a disaster to have more than one instance running. But maybe there is a better way to sample CPU usage than poking tc_Launch/tc_Switch? I have to investigate how Hieronymus profiler works.

Anyway, the task name issue is solved. In internal data structures any invented name can be used when the current Task pointer matches the one being checked.

Also poking the ln_Name field works but it is a hack.

@all

Thanks for the answers.

After some googling, I found a suggestion to just store the old ln_Name and restore it before exit. Seems to work but probably some kittens died in the process.

I was a little bit tired yesterday evening so I try to clarify. My expectation was that if I ran a program called "hello" from shell, it would have appeared with the name "hello" when I walk through the Exec list of tasks (ln_Name). But it didn't.

I tried SetCliProgramName() but as far as I can see, it doesn't affect ln_Name at least. It is always "Shell Process".

Technically there is just one task (or process) with no child tasks (or processes).

If a program is started from shell, is it possible to give it a custom name at runtime using some library API?

@Severin

Regarding Scout, it should be possible to use tc_switch and tc_launch vectors of each task, to calculate their execution time.

@Severin

Good suggestion. Maybe in 0.7?

@ChrisH

In general it would be nice if Exec provided API to query CPU load. Then all these meters would be compatible and much simpler technically.

Exec could also keep book on each task's CPU time while scheduling them.

CPU Watcher has now a GitHub project. Version 0.6 is in the OS4Depot upload queue.

Code was cleaned, some bugs were fixed and it can now use real transparency. Tested on Sam440 only so far so it would be good to hear other results :)

https://github.com/capehill/cpuwatcher

@alfkil

I realize this may not be the answer you are looking for, but you could also try sgit and GitHub/BitBucket if this is your own project.

@MickJT

Yes, documentation is here https://wiki.libsdl.org/SDL_GetKeyboardState

Checkkeys also reports escape pressed/released. But what it should report for text?

Regarding bad_alloc, it's also game issue that it doesn't try-catch the error. However, we have to debug why allocation failed.

@MickJT

Esc might be a bug. I see game uses:

https://sourceforge.net/p/tuxfootball/ ... e/src/gameengine.cpp#l115

SDL_GetKeyboardState returns array that should be indexed with SDL_ScanCodes. So m_quitKey should be SDL_SCANCODE_ESCAPE.

bad_alloc...can you find out the line?

https://github.com/Raul-Ramos/Gauntlet-II another SDL2 app.

@Cap

Heh, just finished Dark Souls 3 like a week ago. Phew.

@tlosm:

Half-life engine looks very interesting. What is the status of the project?

@MickJT:

Thanks!

https://sourceforge.net/projects/tuxfootball/ uses SDL2. We already have older port.

@Kamelito

Since SDL provides abstraction over native API, there is a small function call overhead. I don't have any numbers though.

If you consider CompositeTags features, you can draw faster with native call because there is always one drawcall per "sprite" (or quad) in SDL2 (no sprite batching). Natively you could draw many sprites per one drawcall assuming they use the same texture (sprite sheet).

If you consider OpenGL features, it shouldn't matter if you use SDL or not. SDL is only used to setup the context.

Power of SDL comes from the portability. It's generally very easy to target multiple platforms at once. You can get a simple app running on a new OS in a matter of minutes. It's amazing.

SDL2 is not my port, I am one of the contributors. SDL2 can be linked dynamically or statically. It depends on the developer and end user really doesn't need to care. Also SDL1 was a shared object. Nobody has talked about making an Amiga library so I guess there is no interest, but hey, it's open source.

Personally I don't see a big benefit of .library over .so in this case. If developers link dynamically, binaries are not bigger than with native libraries (why would they be?).

Sadly, you cannot fill modern hard discs with AmigaOS binaries. There is a couple of reasons for that :)

@Raziel

Danke! I hope people will find SDL2 useful :)

I'm trying to fix the remaining known bugs, maybe another release will follow after that.

@MickJT

AmigaOS failed to map the rawkey 59 (-> 0). If there are duplicated scancodes listed in the table, the later ones overwrite the earlier mappings.

By scancode 46 and 48 I mean SDL enumerations. Rawkeys are possibly 0xD and 0x1B (according to the Wiki picture). One left to backspace and one left to enter (upper).

@MickJT

Debugged this problem a bit and it looks like rawkey 59 was unmappable, and because amiga_scancode_table should have only unique items, rawkey 59's mapping (as 0) replaced rawkey 58's mapping which caused "unknown" output in checkkeys for SDL_SCANCODE_SLASH. Please test r126.

Also scancodes 46 and 48 cause unknown but I didn't investigate further yet.

@thellier

Yeah, I suppose there has to be a secondary bitmap allocated if application requests color modulation for a given texture.

So:

1) original texture x*y
2) "modulator" texture (like 4*4)
3) resulting, "modulated" texture x*y, that will be displayed after magic dust falls