PlaneShift

Development => Development Deliberation => Topic started by: yaf on December 25, 2005, 01:32:00 am

Title: not a JPEG file ?
Post by: yaf on December 25, 2005, 01:32:00 am

hi!
I\'m new here but was reading these forums for quite a long time and did this small job... oh, and sorry for my bad English

fortunately I managed to get PlaneShift_CBV0.3.011.linux.x86-2.run (as I have modem connection), and also downloaded patches PlaneShift_CBV0.3.012P-r2.i686.bin and PlaneShift-CBV0.3.0.12-r4-i686.tar.bz2 to upgrade it to 0.3.012. The problem was that psclient from first 2 archives ran practically normal but the one from last archive displayed much more warnings including this:

Not a JPEG file: starts with 0x00...

and no textures were shown in the login window, only some text strings

I\'ve found some threads with this error but all they ended without solution...

I\'m kinda new to Linux and therefore to such things like games under Linux, building sources under Linux, CVS... and even programming (under any OS):D
but I\'ve learned CVS basics and some more tools, downloaded CVS versions of PS for 24th October 2005 and 30th November 2005 (dates for last 2 archives) and compared them. it took me 20 min to find the bug, but overall time is much longer! so, please, dont say its useless job (anyway I\'ve learned many things).

So, u just have to revert file common/paws/pawstexturemanager.cpp from revision 1.16 to 1.15 i.e. uncomment 3 lines that \"doesn\'t seem to have an effect...\"

I also have window$, but since I dont use it so far, I didnt jeck PS under it:rolleyes:

maybe u already know this bug, but than I havent found that in forum
oh, and sorry for such a long... post but anyway I got to learn... (English in particular):baby:

yeah, great game!

Title:
Post by: stfrn on December 26, 2005, 10:57:20 pm

Quote

Originally posted by yaf
So, u just have to revert file common/paws/pawstexturemanager.cpp from revision 1.16 to 1.15 i.e. uncomment 3 lines that \"doesn\'t seem to have an effect...\"

Hm, the only problem with this change is the rest of that comment:

Quote

Comment by dfryer
This doesn\'t seem to have an effect, and crashes some Macs.

Now, keep in mind that by some he meant a signifigant number, and that I have not heard of any linux users currently having that bug- it used to be common in older updates. So in this case it\'s better to stick with the change that fixes unavoidable crashes, and causes avoidable bugs.

A fix that works well for all platforms would be prefered of course.

Title:
Post by: yaf on December 27, 2005, 01:56:26 am

I\'m not sure, but these two threads are about this problem:

http://www.planeshift3d.com/wbboard/thread.php?threadid=20956&boardid=41

http://www.planeshift3d.com/wbboard/thread.php?threadid=21193&boardid=41&styleid=4&page=1#4

So it seems to be very specific and there\'s no solution yet...

But if my suggestion will cause any other bugs... Anyway its up to you to decide whats better, especially since PS still works oddly for me (common problem, but now it works!), as most CS-based games (e.g. VegaStrike is a bit slow).

So good luck...

Title:
Post by: yaf on December 27, 2005, 08:01:04 am

One more (stupid?..) suggestion. If uncommenting those 3 lines will crash only some Mac users (and possibly Windows users) u could use this

Code: [Select]

#ifdef CS_PLATFORM_UNIX
...
#else
...
#endif
(assuming that Linux users didnt have any problem with these lines not commented)

Title:
Post by: stfrn on December 27, 2005, 12:43:28 pm

That\'s already a better solution :) It is never good to just change something back because you do not see the reason it was done.

Title: yaf
Post by: Bereror on December 27, 2005, 01:49:31 pm

Could you try uncommenting these two flags one by one to see, which one is the reason behind the problem you get? Having an #ifdef for Linux clients is a workaround, but not a fix. It might be that there is a bug in the CrystalSpace code that would be nice to find out and report to the developers.

Title:
Post by: yaf on December 27, 2005, 08:28:38 pm

No, no... Both flags must present.
And there seems to be no difference in warnings output when commenting one or other.

Surely, using #ifdef is a temporary solution. And the number of such unlucky Linux users like me   ;(   is not high, so you might won\'t use it at all...

Title:
Post by: yaf on December 27, 2005, 09:12:57 pm

And did anyone tried to figure out why Macs crashed? Its also a bug so there must be some work to fix it. Its not the right place here to ask such question, though, as its bug in CS?
I think maybe fixing Mac-side of the bug will solve this problem too... If you understand what I mean.

Title:
Post by: Bereror on December 27, 2005, 09:39:29 pm

There is something with these two flags that makes some Mac clients to crash and some Linux clients to fail loading textures. The description of these flags does not give an idea, where the problem could be:

Code: [Select]


[COLOR=blue]/**
 * This texture will not be tiled, and color lookups outside the 0..1 range
 * will be clamped to the edge of the texture.
 */[/COLOR]
[COLOR=darkgreen]#define CS_TEXTURE_CLAMP     0x00000010[/COLOR]

[COLOR=blue]/**
 * Store texture as non-power-of-two sized (NPOTS) if possible.
*/[/COLOR]
[COLOR=darkgreen]#define CS_TEXTURE_NPOTS     0x00000040[/COLOR]


PS. Double-posting is not tolerated in these forums. Use the Edit Post button to add new information to your previous post if nobody has replied yet ;)