MantisBT - VCMI
View Issue Details
IDProjectCategoryView StatusDate SubmittedLast Update
0000958VCMISound, music, videospublic2012-05-16 08:232014-05-30 17:41
Reporterdouggie_m 
Assigned ToIvan 
PrioritynormalSeverityminorReproducibilityalways
StatusclosedResolutionfixed 
PlatformPowerPC (Big Endian)OSGNU/LinuxOS VersionDebian Sid
Product Version0.88 
Target VersionFixed in Version0.89 
Summary0000958: Wrong colors in some places on map and inside town screen
DescriptionI built VCMI 0.88 release tarball, extracted the required .lod files and etc. from WoG 3.58f distribution, and:
1. Good news - it works!
2. Bug - it shows wrong colors in some places.

Screenshot is attached.
Steps To ReproduceJust start any map, and you'll see it.
I suppose it is endianess bug, i.e. you may not see it on x86 (endian little architecture).
Additional Information
TagsNo tags attached.
Relationships
Attached Filespng Screenshot - 05162012 - 03:55:10 AM.png (602,045) 2012-05-16 08:23
https://bugs.vcmi.eu/file_download.php?file_id=927&type=bug
png

png Screenshot - 05162012 - 04:32:46 AM.png (1,313,869) 2012-05-16 08:42
https://bugs.vcmi.eu/file_download.php?file_id=928&type=bug
png Screenshot - 05162012 - 10:55:00 AM.png (566,759) 2012-05-16 15:03
https://bugs.vcmi.eu/file_download.php?file_id=929&type=bug
png

png Screenshot - 05162012 - 12:09:17 PM.png (1,004,074) 2012-05-16 16:17
https://bugs.vcmi.eu/file_download.php?file_id=930&type=bug
png

png Screenshot - 05162012 - 03:44:29 PM.png (1,014,023) 2012-05-16 19:49
https://bugs.vcmi.eu/file_download.php?file_id=931&type=bug
png

png Screenshot - 05162012 - 03:57:26 PM.png (535,715) 2012-05-16 19:59
https://bugs.vcmi.eu/file_download.php?file_id=932&type=bug
png

png Screenshot - 05162012 - 04:36:22 PM.png (1,022,548) 2012-05-16 20:38
https://bugs.vcmi.eu/file_download.php?file_id=933&type=bug
png

patch endianless_patch.patch (4,546) 2012-05-16 21:18
https://bugs.vcmi.eu/file_download.php?file_id=934&type=bug
patch endianless_patch_corrected.patch (33,615) 2012-05-16 21:29
https://bugs.vcmi.eu/file_download.php?file_id=935&type=bug
png Screenshot - 05162012 - 05:49:19 PM.png (1,648,970) 2012-05-16 21:51
https://bugs.vcmi.eu/file_download.php?file_id=936&type=bug
png Screenshot - 05172012 - 03:35:32 PM.png (1,325,600) 2012-05-17 19:38
https://bugs.vcmi.eu/file_download.php?file_id=937&type=bug
patch endian-minimap.patch (4,364) 2012-05-17 22:48
https://bugs.vcmi.eu/file_download.php?file_id=938&type=bug
png Screenshot - 05182012 - 05:15:18 AM.png (31,139) 2012-05-18 09:17
https://bugs.vcmi.eu/file_download.php?file_id=939&type=bug
png

png Screenshot - 05212012 - 12:30:10 PM.png (1,355,805) 2012-05-21 16:33
https://bugs.vcmi.eu/file_download.php?file_id=941&type=bug
png Screenshot - 05222012 - 01:02:36 PM.png (1,658,304) 2012-05-22 17:09
https://bugs.vcmi.eu/file_download.php?file_id=943&type=bug
png Screenshot - 05252012 - 03:03:15 PM.png (1,679,065) 2012-05-25 19:11
https://bugs.vcmi.eu/file_download.php?file_id=948&type=bug
Issue History
Date ModifiedUsernameFieldChange
2012-05-16 08:23douggie_mNew Issue
2012-05-16 08:23douggie_mFile Added: Screenshot - 05162012 - 03:55:10 AM.png
2012-05-16 08:34douggie_mNote Added: 0002445
2012-05-16 08:42douggie_mFile Added: Screenshot - 05162012 - 04:32:46 AM.png
2012-05-16 12:15IvanNote Added: 0002447
2012-05-16 12:15IvanAssigned To => Ivan
2012-05-16 12:15IvanStatusnew => assigned
2012-05-16 12:36IvanNote Edited: 0002447bug_revision_view_page.php?bugnote_id=2447#r1594
2012-05-16 13:23douggie_mNote Added: 0002449
2012-05-16 13:49douggie_mNote Edited: 0002449bug_revision_view_page.php?bugnote_id=2449#r1596
2012-05-16 14:06douggie_mNote Edited: 0002449bug_revision_view_page.php?bugnote_id=2449#r1597
2012-05-16 14:59douggie_mNote Added: 0002450
2012-05-16 15:03douggie_mFile Added: Screenshot - 05162012 - 10:55:00 AM.png
2012-05-16 15:06IvanNote Added: 0002451
2012-05-16 15:07IvanNote Edited: 0002451bug_revision_view_page.php?bugnote_id=2451#r1599
2012-05-16 15:15douggie_mNote Added: 0002452
2012-05-16 15:23IvanNote Added: 0002453
2012-05-16 15:34douggie_mNote Added: 0002454
2012-05-16 15:40douggie_mNote Edited: 0002454bug_revision_view_page.php?bugnote_id=2454#r1601
2012-05-16 15:49IvanNote Added: 0002455
2012-05-16 15:51douggie_mNote Edited: 0002454bug_revision_view_page.php?bugnote_id=2454#r1602
2012-05-16 15:51IvanNote Edited: 0002455bug_revision_view_page.php?bugnote_id=2455#r1604
2012-05-16 15:53douggie_mNote Added: 0002456
2012-05-16 15:54douggie_mNote Edited: 0002456bug_revision_view_page.php?bugnote_id=2456#r1606
2012-05-16 15:56IvanNote Edited: 0002455bug_revision_view_page.php?bugnote_id=2455#r1607
2012-05-16 15:58IvanNote Added: 0002457
2012-05-16 15:59douggie_mNote Edited: 0002456bug_revision_view_page.php?bugnote_id=2456#r1608
2012-05-16 16:01IvanNote Edited: 0002455bug_revision_view_page.php?bugnote_id=2455#r1609
2012-05-16 16:02IvanNote Edited: 0002457bug_revision_view_page.php?bugnote_id=2457#r1611
2012-05-16 16:02douggie_mNote Added: 0002458
2012-05-16 16:02douggie_mNote Edited: 0002458bug_revision_view_page.php?bugnote_id=2458#r1613
2012-05-16 16:05douggie_mNote Edited: 0002458bug_revision_view_page.php?bugnote_id=2458#r1614
2012-05-16 16:10douggie_mNote Added: 0002459
2012-05-16 16:17douggie_mFile Added: Screenshot - 05162012 - 12:09:17 PM.png
2012-05-16 16:34IvanNote Added: 0002460
2012-05-16 16:44IvanNote Added: 0002461
2012-05-16 16:48douggie_mNote Added: 0002462
2012-05-16 17:00IvanNote Added: 0002463
2012-05-16 17:22IvanNote Added: 0002464
2012-05-16 17:31IvanNote Edited: 0002464bug_revision_view_page.php?bugnote_id=2464#r1616
2012-05-16 19:27douggie_mNote Added: 0002465
2012-05-16 19:28douggie_mNote Edited: 0002465bug_revision_view_page.php?bugnote_id=2465#r1618
2012-05-16 19:35douggie_mNote Edited: 0002465bug_revision_view_page.php?bugnote_id=2465#r1619
2012-05-16 19:48douggie_mNote Added: 0002466
2012-05-16 19:49douggie_mFile Added: Screenshot - 05162012 - 03:44:29 PM.png
2012-05-16 19:56douggie_mNote Edited: 0002466bug_revision_view_page.php?bugnote_id=2466#r1621
2012-05-16 19:59douggie_mFile Added: Screenshot - 05162012 - 03:57:26 PM.png
2012-05-16 20:04douggie_mNote Edited: 0002466bug_revision_view_page.php?bugnote_id=2466#r1622
2012-05-16 20:06douggie_mNote Edited: 0002466bug_revision_view_page.php?bugnote_id=2466#r1623
2012-05-16 20:07douggie_mNote Edited: 0002466bug_revision_view_page.php?bugnote_id=2466#r1624
2012-05-16 20:12douggie_mNote Edited: 0002466bug_revision_view_page.php?bugnote_id=2466#r1625
2012-05-16 20:14IvanNote Added: 0002467
2012-05-16 20:21douggie_mNote Added: 0002468
2012-05-16 20:30douggie_mNote Added: 0002469
2012-05-16 20:36douggie_mNote Added: 0002470
2012-05-16 20:38douggie_mFile Added: Screenshot - 05162012 - 04:36:22 PM.png
2012-05-16 20:53douggie_mNote Edited: 0002470bug_revision_view_page.php?bugnote_id=2470#r1627
2012-05-16 21:17douggie_mNote Added: 0002471
2012-05-16 21:18douggie_mFile Added: endianless_patch.patch
2012-05-16 21:29douggie_mNote Edited: 0002471bug_revision_view_page.php?bugnote_id=2471#r1629
2012-05-16 21:29douggie_mFile Added: endianless_patch_corrected.patch
2012-05-16 21:51douggie_mFile Added: Screenshot - 05162012 - 05:49:19 PM.png
2012-05-16 21:52douggie_mNote Edited: 0002471bug_revision_view_page.php?bugnote_id=2471#r1630
2012-05-16 21:56douggie_mNote Added: 0002472
2012-05-16 22:20douggie_mNote Added: 0002473
2012-05-16 22:34douggie_mNote Edited: 0002473bug_revision_view_page.php?bugnote_id=2473#r1632
2012-05-16 22:39douggie_mNote Edited: 0002473bug_revision_view_page.php?bugnote_id=2473#r1633
2012-05-16 22:47douggie_mNote Deleted: 0002472
2012-05-16 23:01douggie_mNote Edited: 0002473bug_revision_view_page.php?bugnote_id=2473#r1634
2012-05-17 10:48IvanNote Added: 0002474
2012-05-17 19:08douggie_mNote Added: 0002478
2012-05-17 19:36douggie_mNote Edited: 0002478bug_revision_view_page.php?bugnote_id=2478#r1636
2012-05-17 19:38douggie_mFile Added: Screenshot - 05172012 - 03:35:32 PM.png
2012-05-17 19:39douggie_mNote Edited: 0002478bug_revision_view_page.php?bugnote_id=2478#r1637
2012-05-17 19:42douggie_mNote Edited: 0002478bug_revision_view_page.php?bugnote_id=2478#r1638
2012-05-17 20:22douggie_mNote Added: 0002479
2012-05-17 20:50IvanNote Added: 0002480
2012-05-17 21:19douggie_mNote Added: 0002481
2012-05-17 21:19douggie_mNote Edited: 0002473bug_revision_view_page.php?bugnote_id=2473#r1639
2012-05-17 21:35IvanNote Added: 0002482
2012-05-17 21:52douggie_mNote Added: 0002483
2012-05-17 21:57douggie_mNote Added: 0002484
2012-05-17 22:00douggie_mNote Edited: 0002484bug_revision_view_page.php?bugnote_id=2484#r1641
2012-05-17 22:48IvanFile Added: endian-minimap.patch
2012-05-17 22:48IvanNote Added: 0002485
2012-05-18 08:51douggie_mNote Edited: 0002478bug_revision_view_page.php?bugnote_id=2478#r1642
2012-05-18 09:03douggie_mNote Added: 0002486
2012-05-18 09:05douggie_mNote Edited: 0002486bug_revision_view_page.php?bugnote_id=2486#r1644
2012-05-18 09:16douggie_mNote Added: 0002487
2012-05-18 09:17douggie_mFile Added: Screenshot - 05182012 - 05:15:18 AM.png
2012-05-18 09:26douggie_mNote Edited: 0002487bug_revision_view_page.php?rev_id=1646
2012-05-18 09:27douggie_mNote Edited: 0002487bug_revision_view_page.php?rev_id=1647
2012-05-18 09:51douggie_mNote Deleted: 0002487
2012-05-18 10:15douggie_mNote Added: 0002488
2012-05-18 10:29IvanNote Added: 0002489
2012-05-18 10:55IvanNote Added: 0002490
2012-05-18 11:13douggie_mNote Added: 0002491
2012-05-18 11:23douggie_mNote Edited: 0002491bug_revision_view_page.php?bugnote_id=2491#r1649
2012-05-18 17:37IvanNote Added: 0002493
2012-05-18 17:37IvanStatusassigned => resolved
2012-05-18 17:37IvanFixed in Version => 0.89
2012-05-18 17:37IvanResolutionopen => fixed
2012-05-21 16:29douggie_mNote Added: 0002515
2012-05-21 16:33douggie_mFile Added: Screenshot - 05212012 - 12:30:10 PM.png
2012-05-22 17:06douggie_mNote Edited: 0002515bug_revision_view_page.php?bugnote_id=2515#r1657
2012-05-22 17:07douggie_mNote Added: 0002522
2012-05-22 17:07douggie_mStatusresolved => feedback
2012-05-22 17:07douggie_mResolutionfixed => reopened
2012-05-22 17:09douggie_mFile Added: Screenshot - 05222012 - 01:02:36 PM.png
2012-05-22 17:26IvanNote Added: 0002525
2012-05-22 17:42douggie_mNote Added: 0002526
2012-05-22 17:42douggie_mStatusfeedback => assigned
2012-05-22 17:43douggie_mNote Edited: 0002526bug_revision_view_page.php?bugnote_id=2526#r1659
2012-05-25 11:58IvanNote Added: 0002550
2012-05-25 12:57majaczekNote Added: 0002552
2012-05-25 19:11douggie_mFile Added: Screenshot - 05252012 - 03:03:15 PM.png
2012-05-25 19:11douggie_mNote Added: 0002560
2012-05-25 19:44IvanNote Added: 0002561
2012-05-25 19:44IvanStatusassigned => resolved
2012-05-25 19:44IvanResolutionreopened => fixed
2012-05-25 19:47douggie_mNote Added: 0002562
2014-05-30 17:41beegeeStatusresolved => closed

Notes
(0002445)
douggie_m   
2012-05-16 08:34   
Attaching another screenshot (showing the town screen).
(0002447)
Ivan   
2012-05-16 12:15   
(edited on: 2012-05-16 12:36)
Looks like endianess bug indeed. Will check.

This is not required but will be helpful. Can you compile VCMI using latest sources from SVN? It won't fix bug right now but I have no other way to check how my fix will work.
(0002449)
douggie_m   
2012-05-16 13:23   
(edited on: 2012-05-16 14:06)
> Can you compile VCMI using latest sources from SVN?
Why not. I tried (current) revision 2682 and this issue is still present.
(0002450)
douggie_m   
2012-05-16 14:59   
I did commented out the following in client/CbitmapHandler.cpp:84

#if 0 /* (SDL_BYTEORDER == SDL_BIG_ENDIAN) */
                        tp.b = pcx[it++];
                        tp.g = pcx[it++];
                        tp.r = pcx[it++];
#else
                        tp.r = pcx[it++];
                        tp.g = pcx[it++];
                        tp.b = pcx[it++];
#endif

and some parts started to be painted correctly (like chosen hero's portrait, bottom resources line, in-game message dialogs, etc.). Screenshot will follow.
(0002451)
Ivan   
2012-05-16 15:06   
(edited on: 2012-05-16 15:07)
Yup. Found that one too - it should fix all blue-ness.

Yellow-ness most possibly is coming from client/SDL_Extensions:
Here we have several ColorPutter::putColor methods.
Problem is: color with RGB = 222, 243, 255 (Font that supposed to be white) after blitting turns into RGB = 243, 255, 0 (First component missing)

And there is one more separate issue - minimap.
(0002452)
douggie_m   
2012-05-16 15:15   
Well, SDL says:
> Warning: SDL says that 24bpp is wrong and suggests 32

That zero comes from alpha (or "unused"), I suppose.
(0002453)
Ivan   
2012-05-16 15:23   
It means that there is mismatch between X color resolution and in-game one. Shouldn't affect game in any way.
If you wish to remove that warning open file ~/.vcmi/config/settings.json and add this line in "video" section:
"bitsPerPixel" : 32,

Ok. Bug is in ColorPutter::PutColor method in SDL_Extensions.h:238
Now looking for a way to fix...
(0002454)
douggie_m   
2012-05-16 15:34   
(edited on: 2012-05-16 15:51)
I'm testing with

        else if(incrementPtr == -1)
        {
                ptr -= 4;
                ptr[0] = B;
                ptr[1] = G;
                ptr[2] = R;
                if (bpp == 4) ptr[3] = 0;

No, this was a bad idea ;)

This variant:

                if (bpp == 4) {
                        ptr -= 4;
                        ptr[0] = B;
                        ptr[1] = G;
                        ptr[2] = R;
                        ptr[3] = 0;
                } else {
                        ptr -= 3;
                        ptr[0] = B;
                        ptr[1] = G;
                        ptr[2] = R;
                }

Does produce the same result as original.
(0002455)
Ivan   
2012-05-16 15:49   
(edited on: 2012-05-16 16:01)
Try replacing that method with this code:
{
    Uint8 *ptr2 = ptr;
    if(incrementPtr == 0)
    {
        if(bpp == 4)
            *ptr2++ = 0;

        ptr2[0] = R;
        ptr2[1] = G;
        ptr2[2] = B;
    }
    else if(incrementPtr == 1)
    {
        if(bpp == 4)
            *ptr2++ = 0;

        *ptr2++ = R;
        *ptr2++ = G;
        *ptr2++ = B;
        ptr = ptr2;
    }
    else if(incrementPtr == -1)
    {
        *(--ptr2) = B;
        *(--ptr2) = G;
        *(--ptr2) = R;
        
        if(bpp == 4)
            *(--ptr2) = 0;
        ptr = ptr2;
    }
    else
    {
        assert(0);
    }
}
(0002456)
douggie_m   
2012-05-16 15:53   
(edited on: 2012-05-16 15:59)
> First byte is unused\alpha ... After this - 3 RGB bytes

That means that the color format is ARGB and not RGBA?

> Try replacing that method with this code

Okay, trying it now.

By the way, shouldn't we restore `ptr' back after incrementing it?
(0002457)
Ivan   
2012-05-16 15:58   
(edited on: 2012-05-16 16:02)
Looks to be so. That's the only way I found to get such image

Note: I've edited above code. Previous version could crash.

Yes. ptr should be restored if increment != 0. My bad.
(0002458)
douggie_m   
2012-05-16 16:02   
(edited on: 2012-05-16 16:05)
Your new version doesn't increment/decrement for +1/-1 cases.

I'm trying previous one with

    if(incrementPtr == 0)
    {
        if(bpp == 4)
            *ptr++ = 0;

        ptr[0] = R;
        ptr[1] = G;
        ptr[2] = B;
        
        if(bpp == 4)
            ptr--;
    }

UPD: You fixed it one more time ;) These edits make me crazy.
(0002459)
douggie_m   
2012-05-16 16:10   
Very nice.

Now yep, minimap left.
(0002460)
Ivan   
2012-05-16 16:34   
Sorry. Will try less edits next time :)

It seems that in all cases (map, shadows, button frame) blue component is incorrect. Looking into this issues right now.
(0002461)
Ivan   
2012-05-16 16:44   
Ok. Previous two methods: lines 207 and 229
1) Add at start:
Uint8 * ptr2 = ptr;
if (bpp == 4)
    ptr2++;

2) Replace all access to ptr as array to ptr2:
ptr[0] -> ptr2[0]
ptr[1] -> ptr2[1]
ptr[1] -> ptr2[2]

No need to update ptr after this.

This should fix at least some issues.

When everything will be fixed I'll make better patch that will work on both little and big-endian systems.
(0002462)
douggie_m   
2012-05-16 16:48   
> It seems that in all cases (map, shadows, button frame) blue component is incorrect. Looking into this issues right now.

On the last screenshot map and shadows look nice for me. Button frame is little emo-ish, yeah.

By the way, can you upload screenshot to see how it *should* look like?
(0002463)
Ivan   
2012-05-16 17:00   
By "map" I meant "minimap"

http://forum.vcmi.eu/album.php [^] - VCMI screenshots, some may be outdated
http://dl.dropbox.com/u/22372764/vcmi/vcmi-current.png [^] - my screenshot of this map
(0002464)
Ivan   
2012-05-16 17:22   
(edited on: 2012-05-16 17:31)
This is what I have right now:
4 modified ColorPutter methods, lines 205-285:

Sorry. Missed few ptr2's

http://pastebin.com/9j1WJfUt [^]
(0002465)
douggie_m   
2012-05-16 19:27   
(edited on: 2012-05-16 19:35)
> error: ‘Uint’ was not declared in this scope
(I changed them to Uint8)

And please, use `diff -u' next time to easily feed that to `patch'.

UPD: I used your patch, but minimap is still the same.
(0002466)
douggie_m   
2012-05-16 19:48   
(edited on: 2012-05-16 20:12)
Look what I've found, CAdvmapInterface.cpp:

--- client/CAdvmapInterface.cpp.orig 2012-05-16 15:36:39.248625059 -0400
+++ client/CAdvmapInterface.cpp 2012-05-16 15:42:35.787735261 -0400
@@ -77,16 +77,16 @@
                vinya.first = m["terrain_id"].Float();
 
                const JsonVector &unblocked_vec = m["unblocked"].Vector();
- vinya.second.r = unblocked_vec[0].Float();
- vinya.second.g = unblocked_vec[1].Float();
- vinya.second.b = unblocked_vec[2].Float();
+ vinya.second.r = 255; //unblocked_vec[2].Float();
+ vinya.second.g = 255; //unblocked_vec[1].Float();
+ vinya.second.b = 255; //unblocked_vec[0].Float();
                vinya.second.unused = 255;
                colors.insert(vinya);
 
                const JsonVector &blocked_vec = m["blocked"].Vector();
- vinya.second.r = blocked_vec[0].Float();
- vinya.second.g = blocked_vec[1].Float();
- vinya.second.b = blocked_vec[2].Float();
+ vinya.second.r = 255; //blocked_vec[2].Float();
+ vinya.second.g = 255; //blocked_vec[1].Float();
+ vinya.second.b = 255; //blocked_vec[0].Float();
                vinya.second.unused = 255;
                colorsBlocked.insert(vinya);
        }

I first tried to swap rgb <-> bgr, but this gave me no better results. But setting them to 0xFF makes minimap be white, except castles, heroes, and emo ;) border. Maybe this can help. Screenshot follows.

UPD: `vinya.second.r = 255; vinya.second.g = 0; vinya.second.b = 0;' makes minimap the same emo magenta as the border (and *not red*).
UPD2: green=255, r,b=0 gives cyan color; blue=255, r,g=0 gives... no, not yellow, but clean blue! r,g,b=0 gives the same blue too.

Note: magenta is R=FF, G=0, B=FF; cyan is R=0, G=FF, B=FF.
(0002467)
Ivan   
2012-05-16 20:14   
Looks to be same problem as yellowness. Will try to make sense of it later - it's quite late here.

Tomorrow I'll finish "endian-friendly" patch and will re-check every place with direct access to images. If I won't miss anything then all you'll need to do is to re-check subpixels order.

One question - have you set bits per pixel in config to 32? Or you're using default value (24)?
(0002468)
douggie_m   
2012-05-16 20:21   
> have you set bits per pixel in config to 32? Or you're using default value (24)?
I'm following the rule "Warnings can be ignored" ;) So I haven't change anything.
(0002469)
douggie_m   
2012-05-16 20:30   
I found how it is drawn, it is "SDL_PutPixelWithoutRefresh", part of the same "SDL Extensions":

void CSDL_Ext::SDL_PutPixelWithoutRefresh(SDL_Surface *ekran, const int & x, const int & y, const Uint8 & R, const Uint8 & G, const Ui$
{
        Uint8 *p = getPxPtr(ekran, x, y);
        getPutterFor(ekran, false)(p, R, G, B);

        //needed?
        if(ekran->format->BytesPerPixel==4)
                p[3] = A;
}
(0002470)
douggie_m   
2012-05-16 20:36   
(edited on: 2012-05-16 20:53)
Bingo! It is "not needed" actually. Just commenting out

if(ekran->format->BytesPerPixel==4)
                p[3] = A;

gave me cool minimap!

UPD: By the way, I know why other SDL-games are blueish now ;) They do use these SDL_Extensions, which seems to be popular. Is there devs for them? If so, these changes shall be reported to them.
(0002471)
douggie_m   
2012-05-16 21:17   
(edited on: 2012-05-16 21:52)
Attaching patch against current SVN (2684).

UPD: oops

UPD2: attached screenshot with the game built with endianless_patch_corrected.patch applied to fresh svn; game is 100% playable now

Great work! Thanks, Ivan.
(0002473)
douggie_m   
2012-05-16 22:20   
(edited on: 2012-05-17 21:19)
OOPS, I started to actually play a game, but... It does segfault on each battle! Is this known issue? Shall be reported separately?

--------------------------------
Warning: empty path found...
Segmentation fault
Server finished
Lost connection to server, ending listening thread!
read: End of file
Something wrong, lost connection while game is still ongoing...
terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::system::system_error> >'
  what(): read: End of file
Aborted
--------------------------------

GDB gives me:

--------------------------------
Program received signal SIGABRT, Aborted.
[Switching to Thread 0xec50b450 (LWP 12163)]
0x0de9098c in *__GI_raise (sig=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:64
64 ../nptl/sysdeps/unix/sysv/linux/raise.c: No such file or directory.
(gdb) bt
#0 0x0de9098c in *__GI_raise (sig=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:64
#1 0x0de96040 in *__GI_abort () at abort.c:92
0000002 0x0e16a9d0 in __gnu_cxx::__verbose_terminate_handler() () from /usr/lib/powerpc-linux-gnu/libstdc++.so.6
0000003 0x0e1679e8 in ?? () from /usr/lib/powerpc-linux-gnu/libstdc++.so.6
0000004 0x0e167a2c in std::terminate() () from /usr/lib/powerpc-linux-gnu/libstdc++.so.6
0000005 0x0e167ed8 in __cxa_rethrow () from /usr/lib/powerpc-linux-gnu/libstdc++.so.6
0000006 0x1035319c in CClient::run (this=0x1119d0f0) at Client.cpp:162
0000007 0x10448c30 in boost::_mfi::mf0<void, CClient>::operator() (this=0x111a3574, p=0x1119d0f0)
    at /usr/include/boost/bind/mem_fn_template.hpp:49
0000008 0x10448a38 in boost::_bi::list1<boost::_bi::value<CClient*> >::operator()<boost::_mfi::mf0<void, CClient>, boost::_bi::list0> (
    this=0x111a357c, f=..., a=...) at /usr/include/boost/bind/bind.hpp:253
0000009 0x104488d8 in boost::_bi::bind_t<void, boost::_mfi::mf0<void, CClient>, boost::_bi::list1<boost::_bi::value<CClient*> > >::operator() (this=0x111a3574) at /usr/include/boost/bind/bind_template.hpp:20
0000010 0x10447cfc in boost::detail::thread_data<boost::_bi::bind_t<void, boost::_mfi::mf0<void, CClient>, boost::_bi::list1<boost::_bi::value<CClient*> > > >::run (this=0x111a3470) at /usr/include/boost/thread/detail/thread.hpp:62
0000011 0x0e49c40c in ?? () from /usr/lib/libboost_thread.so.1.49.0
0000012 0x0de247b0 in start_thread (arg=<optimized out>) at pthread_create.c:306
0000013 0x0df4ab10 in clone () at ../sysdeps/unix/sysv/linux/powerpc/powerpc32/clone.S:123
--------------------------------

UPD: it does actually look like bug in either GLibC or Boost.
UPD2: to double-recheck, I built fresh svn^W 0.88 release w/o this patche, ignoring colors issue; and I was able to play battles.
(0002474)
Ivan   
2012-05-17 10:48   
I'm back.

SDL_Extension is part of VCMI but I suppose most of SDL-based projects have similar code. SDL itself provides only basic functionality.

You patch won't work as it because it will mess up on little-endian systems.
I've applied all your fixes in endian-independent way - update to rev 2685. It should work without any glitches and without any additional patches.

Some areas to re-check:
- shadows. They should be gray, not green as on your screenshots
- puzzle map. Find some obelisks on map - opened areas should look like old photo (sepia effect): http://forum.vcmi.eu/album_pic.php?pic_id=39 [^]
- battles.
- town and town fort windows.
(0002478)
douggie_m   
2012-05-17 19:08   
(edited on: 2012-05-18 08:51)
> I've applied all your fixes in endian-independent way - update to rev 2685. It should work without any glitches and without any additional patches.
Great! I checked out and is ready to build fresh revision 2686.

Hang about a half of hour for the test report ;)

> - shadows. They should be gray, not green as on your screenshots
--They are green.
++Shadows are gray now. Neutral-owned stuff on minimap is still green with r2686.
UPD: By the way, owned-by-player minimap's stuff is black, not red (as on your screenshots).

> - puzzle map
Looks fine. See screenshot.

> - battles
Crash.
terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::system::system_error> >'

> - town and town fort windows.
They are all fine.
(0002479)
douggie_m   
2012-05-17 20:22   
> SDL_Extension is part of VCMI but I suppose most of SDL-based projects have similar code. SDL itself provides only basic functionality.

Ah, okay. At least, I know where to start to dig [spoiler]for the Grail[/spoiler] ;)
(0002480)
Ivan   
2012-05-17 20:50   
Nice! Screenshot looks fine - even shadows on snow are gray as they should.
Not sure what's wrong with minimap. Will check.

Regarding crash - this looks like crash on server.
Before starting a battle attach debugger to vcmiserver process. This should show real location of the crash - system_error/ End of file usually means that client lost connection with server.
(0002481)
douggie_m   
2012-05-17 21:19   
Regarding crash on battle - the last working one I tried was 0.88 tarball actually, not svn.
I can try git-svn and then git bisect.
(0002482)
Ivan   
2012-05-17 21:35   
git bisect may help. Or may not - some of our commits are quite large. Stacktrace from server should be more easier to check.

And I think I fixed minimap issue. Will commit fix a bit later.
(0002483)
douggie_m   
2012-05-17 21:52   
> Stacktrace from server should be more easier to check.

Okay, you convicted me to spend some time with gdb ;)
(0002484)
douggie_m   
2012-05-17 21:57   
(edited on: 2012-05-17 22:00)
> I think I fixed minimap issue. Will commit fix a bit later.

Can you post the patch here, just to test it before you apply it to upstream?..
(0002485)
Ivan   
2012-05-17 22:48   
Patch uploaded. I've cut it from my own changes but it should be complete.
(0002486)
douggie_m   
2012-05-18 09:03   
(edited on: 2012-05-18 09:05)
This does definitely look like another bug, not "minor" and not "sound, music, video":

----------------------------------------
$ ps -A | grep vcmiserver
 2309 pts/4 00:00:01 vcmiserver
...
(gdb) attach 2309
...
(gdb) c
...
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0xf6eff450 (LWP 2500)]
0x0f7fdc48 in BattleInfo::setupBattle (tile=..., terrain=3, terType=11,
    armies=0x104285b8, heroes=0x104285c0, creatureBank=false, town=0x0)
    at BattleState.cpp:1753
1753 creatureBank ? commanderBank[i] : commanderField[i]);
(gdb) bt
#0 0x0f7fdc48 in BattleInfo::setupBattle (tile=..., terrain=3, terType=11,
    armies=0x104285b8, heroes=0x104285c0, creatureBank=false, town=0x0)
    at BattleState.cpp:1753
#1 0x0f90fc90 in CGameState::setupBattle (this=0x10840e80, tile=...,
    armies=0x104285b8, heroes=0x104285c0, creatureBank=false, town=0x0)
    at CGameState.cpp:787
0000002 0x10172a50 in CGameHandler::setupBattle (this=0x108408e0, tile=...,
    armies=0x104285b8, heroes=0x104285c0, creatureBank=false, town=0x0)
    at CGameHandler.cpp:1464
0000003 0x101697a4 in CGameHandler::startBattle(CArmedInstance const**, int3, CGHeroInstance const**, bool, boost::function<void (BattleResult*)>, CGTownInstance const*) (this=0x108408e0, armies=0x104285b8, tile=...,
    heroes=0x104285c0, creatureBank=false, cb=..., town=0x0)
    at CGameHandler.cpp:372
0000004 0x10295774 in boost::_mfi::mf6<void, CGameHandler, CArmedInstance const**, int3, CGHeroInstance const**, bool, boost::function<void (BattleResult*)>, CGTownInstance const*>::operator()(CGameHandler*, CArmedInstance const**, int3, CGHeroInstance const**, bool, boost::function<void (BattleResult*)>, CGTownInstance const*) const (this=0x10958bd4, p=0x108408e0, a1=0x104285b8,
    a2=..., a3=0x104285c0, a4=false, a5=..., a6=0x0)
    at /usr/include/boost/bind/mem_fn_template.hpp:732
0000005 0x10295334 in boost::_bi::list7<boost::_bi::value<CGameHandler*>, boost::_bi::value<CArmedInstance const**>, boost::_bi::value<int3>, boost::_bi::value<CGHeroInstance const**>, boost::_bi::value<bool>, boost::_bi::value<boost::function<void (BattleResult*)> >, boost::_bi::value<CGTownInstance const*> >::operator()<boost::_mfi::mf6<void, CGameHandler, CArmedInstance const**, int3, CGHeroInstance const**, bool, boost::function<void (BattleResult*)>, CGTownInstance const*>, boost::_bi::list0>(boost::_bi::type<void>, boost::_mfi::mf6<void, CGameHandler, CArmedInstance const**, int3, CGHeroInstance const**, bool, boost::function<void (BattleResult*)>, CGTownInstance const*>&, boost::_bi::list0&, int) (this=0x10958bdc, f=..., a=...)
    at /usr/include/boost/bind/bind.hpp:670
0000006 0x10295144 in boost::_bi::bind_t<void, boost::_mfi::mf6<void, CGameHandler, CArmedInstance const**, int3, CGHeroInstance const**, bool, boost::function<void (BattleResult*)>, CGTownInstance const*>, boost::_bi::list7<boost::_bi::value<CGameHandler*>, boost::_bi::value<CArmedInstance const**>, boost::_bi::value<int3>, boost::_bi::value<CGHeroInstance const**>, boost::_bi::value<bool>, boost::_bi::value<boost::function<void (BattleResult*)> >, boost::_bi::value<CGTownInstance const*> > >::operator()() (this=0x10958bd4)
    at /usr/include/boost/bind/bind_template.hpp:20
0000007 0x10292858 in boost::detail::thread_data<boost::_bi::bind_t<void, boost::_mfi::mf6<void, CGameHandler, CArmedInstance const**, int3, CGHeroInstance const**, bool, boost::function<void (BattleResult*)>, CGTownInstance const*>, boost::_bi::list7<boost::_bi::value<CGameHandler*>, boost::_bi::value<CArmedInstance const**>, boost::_bi::value<int3>, boost::_bi::value<CGHeroInstance const**>, boost::_bi::value<bool>, boost::_bi::value<boost::function<void (BattleResult*)> >, boost::_bi::value<CGTownInstance const*> > > >::run() (
    this=0x10958ad0) at /usr/include/boost/thread/detail/thread.hpp:62
0000008 0x0f25d40c in ?? () from /usr/lib/libboost_thread.so.1.49.0
0000009 0x0f21c7b0 in start_thread (arg=<optimized out>) at pthread_create.c:306
0000010 0x0ed3fb10 in clone ()
    at ../sysdeps/unix/sysv/linux/powerpc/powerpc32/clone.S:123
(0002488)
douggie_m   
2012-05-18 10:15   
Oops, I suffered Debian Sid power to break everything:

----------------------------------------
Cannot establish connection! Retrying within 2 seconds
Establishing connection...
Found endpoints:
        0: 127.0.0.1:3030
Trying connection to 127.0.0.1:3030 (0)
Problem with connecting:
system:101
Something went wrong... checking for error info
system:101
----------------------------------------
$ ping 127.0.0.1
connect: Network is unreachable
----------------------------------------

So, please, ignore the deleted comment and the last screenshot.
Writing from OS X now (by the way, where's VCMI for OS X? ;)
I dunno how and when I'll fix this no-networking-at-all issue.
(0002489)
Ivan   
2012-05-18 10:29   
VCMI for OS X? We don't have anybody with this OS but as far as I know it works. You need to compile it yourself though.
One note - you'll need at least gcc-4.5. clang or earlier gcc versions won't work

Looks like commanders issue. That also explains why it as not present in 0.88. Open new issue in Battles/Mechanics if you wish
(0002490)
Ivan   
2012-05-18 10:55   
Ok. I think i found it. You need to update configuration files:
copy them (or create symlink) - files in
/usr/local/share/vcmi/config
should be the same as in
/path/to/source/config
(0002491)
douggie_m   
2012-05-18 11:13   
(edited on: 2012-05-18 11:23)
> Ok. I think i found it. You need to update configuration files
If this is about vcmiserver's issue, I've already open a separate issue.
http://bugs.vcmi.eu/view.php?id=960 [^]
(0002493)
Ivan   
2012-05-18 17:37   
Moving to resolved - minimap was fixed in rev 2688.
(0002515)
douggie_m   
2012-05-21 16:29   
(edited on: 2012-05-22 17:06)
-- I do confirm the fix.
++ I confirmed the fix when I applied endian-minimap.patch to r2684. Fresh SVN (r2704) now shows greenish minimap, no magenta border now :(
Screenshot will follow.
(0002522)
douggie_m   
2012-05-22 17:07   
I confirmed the fix when I applied endian-minimap.patch to r2684. Fresh SVN (r2704) now shows greenish minimap.
(0002525)
Ivan   
2012-05-22 17:26   
I won't have time for VCMI for several days but I'll check what went wrong here as soon as I can.
(0002526)
douggie_m   
2012-05-22 17:42   
(edited on: 2012-05-22 17:43)
By the way, if you'll look on Screenshot - 05212012 - 12:30:10 PM.png (it was a result of applying endian-minimap.patch to r2684), you'll not see colored towns and mines on minmap. In current SVN (r2704) they are visible, but the whole minimap is colored wrong, and it is in another "direction" (not blueish, but greenish).
(0002550)
Ivan   
2012-05-25 11:58   
For some reason red channel was set to zero.
Check rev. 2709 - not sure what's causing the glitch so this is mostly guessing "what may be wrong".
(0002552)
majaczek   
2012-05-25 12:57   
24bit color putter was accessed apparently to put unused/alfa field (I read the cod but I don't remember where it actually appear). That means it owerwrite next pixel first color - you need not to use alpha/unused channel in any non-32-bit color putter (maybe use static_assert() to make sure it is never used where "it doesn't exist" ?)
(0002560)
douggie_m   
2012-05-25 19:11   
Well. It. Is. Okay.

You. did. It. Great.

RESOLVED FIXED

P.S. Apparently, the next release version shall be numbered 1.0 or at least 0.9 (but not 0.89).
(0002561)
Ivan   
2012-05-25 19:44   
So it looks that I'm good with guessing.
Moving to resolved.

But why next version should be not 0.89?
(0002562)
douggie_m   
2012-05-25 19:47   
> why next version should be not 0.89?

Well, you did great work. That's why.

Also, these commanders. By the way, will be any option to turn these "WoG" extensions off maybe? To not to get commanders on battle, for example.