Release VCMI - Heroes of Might & Magic III

[ptitSeb]

New build on the repo.
First, this time, the sources are the latest from github, so it's a fresh (cross-compiled) build.
I tweaked the SDL2 renderer creation so the texture it use for its "screen" is a 16bits RGB565 (instead of 32bit RGBA), so Texture Streaming can be used. Because VCMI doesn't really use SDL2 renderer, but simply blit a "screen" that is blitted the old-fashion way, this speed up screen refresh.
Also, because this is basicaly still a SDL1.2 app, it doesn a lot of Blit, so I ported the notaz's NEON Asm Blitter from SDL 1.2 to SDL2.
All in all, this version should be roughly 2* faster then previous one (graphics alone should be at least 3* faster, but there is more than graphics in this game)

Build 06
-----------

• Using latest github source
• Tweaked to use 16bits framebuffer and Texture Streaming
• New SDL2 with notaz's NEON blit from SDL1.2

 

[Linux-SWAT]

Awesome !

 

[asimov-solensan]

Will test and tell the results. Thanks a lot.

 

[asimov-solensan]

After playing for an hour or so I cant tell that there's quite difference. But still on the map it feels very choppy, and this is a GHZ pandora running at 1200.

Besides that there is another thing that I find annoying. In the combat screen (I think that in map also happens) mouse pointer appears in a different position when there is any animation running, just the pointer, because really the mouse itself hasn't change it's position. I don't know if explained myself.

I guess VCMI isn't focused on performance because the requirements for original game are way below what pandora offers. So far I consider the game playable.

I really appreciate the job done by VCMI team and ptitSeb

 

[ptitSeb]

Thanks for the test and the feedback!

About map, I have noticed some strange hickup when moving yes. I'm not sure what cause those pause. I don't think it's something about the drawing this time, and I'm not sure it will be easy to track down.
The mouse issue, I haven't noticed, but I'm using a mouse, so I didn't paid much attention to mouse position, I'll try to check that (maybe it's a side effect of the screen resizing?).

 

[Magic Sam]

Hi all,

@ptitSeb : your latest version doesn't start on my end:

[ START ]--- Starting the application ( runscript.sh ) ----------
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-samuel'
WSEGL_InitialiseDisplay: DRI2OpenConnection failed
Qt: Session management error: Authentication Rejected, reason : None of the a
uthentication protocols specified are supported and host-based authentication
failed
./runscript.sh: line 65: 2806 Segmentation fault ./vcmilauncher
[ FAILED]--- Starting the application ( runscript.sh ) ----------

What am I doing wrong ?

Cheers, Magic Sam

 

[ptitSeb]

I don't know. ckblackm has the same issue.
I have to retry wihout codeblocks PND mounted, and without any data to see if I can reproduce it.

 

[Magic Sam]

@ptitSeb : I tried again from a clean slate, re-downloaded the PND, removed all data but it still didn't start. I then tried with Code::Blocks mounted, and it worked

Hope that helps !

Cheers, Magic Sam

 

[ptitSeb]

@ptitSeb : I tried again from a clean slate, re-downloaded the PND, removed all data but it still didn't start. I then tried with Code::Blocks mounted, and it worked

Hope that helps !

Cheers, Magic Sam

Well, what I guessed, some hidden dependancy Not always easy to track. Hopefulle some "lsof | grep vcmilau | grep codeblocks" will be enough
[doublepost=1523045915,1523039175][/doublepost]So, after many tries, I finaly got a working pnd.
vcmilauncher now launch, and vcmiserver is not broken...

Build 07
-----------

• New repackaged, everything seems to works now

 

[asimov-solensan]

The second map I'm trying is crashing all the time to the desktop. It may be due to swap? I didn't crash when I began playing.

Here's the .out in case it shows something.

 

[ptitSeb]

The second map I'm trying is crashing all the time to the desktop. It may be due to swap? I didn't crash when I began playing.

Here's the .out in case it shows something.

Looks like a engine bug to me?