[Archived] I want to help


Don’t know where the best place to post this so I’ll do it here.
I’ve noticed that on town screen left-bottom info panel doesn’t behave like in H3 and it looks that I fixed this as well as some connected issues:
-all elements now response like in H3 on mouse moves\clicks
-town status bar was not cleaned when nothing is selected
-growth calculation was a bit wrong (in case of town with grail)
-in H3 legion statue effect does not include effect of its components - only global +50%

I’ve attached .diff - I hope everything is OK, never used svn before
townscreen.diff.zip (6.17 KB)


Great job!
I’ve committed it with minor fixes as r1400.
Many thanks for contribution :smiley:

Btw, don’t you want direct SVN commit access?

Some remarks:

if (fort) delete fort;

“if” not needed, delete is guaranteed to do nothing when called with NULL pointer

void CCastleInterface::CTownInfo::activate()
void CCastleInterface::CTownInfo::deactivate()

Not necessary, can be replaced with:

CCastleInterface::CTownInfo::CTownInfo(int BID)
	//rest of c-tor

That possibility has been introduced recently, so most of GUI still don’t use it but new code should. Much cleaner.

if (town->builtBuildings.find(bid)!=town->builtBuildings.end())

FYI I hate typing the container twice, so I made simplier contains function: vstd::contains(town->builtBuildings, bid) (expands to what you’ve written) :wink:

	else if ( bid == 6 )//no fort

Out of range: No fort picture has index 3.

for (bid=10;((ci->town->builtBuildings.find(bid+1)!=ci->town->builtBuildings.end()) && bid<13);bid++) {}	

bid = 10 + ci->town->hallLevel();
(and similar goes for the fort with fortLevel())

  • one occurrence of VLC in CastleInterface causing compile error -> replaced with CGI.


It does for me, though. With a hero with Statue of Legion in a town I get a Statue of Legion as well as a X of Legion bonus to the creatures.


Strange. I tested in WoG (installed over H3Complete), with a hero with Statue of Legion in town (at gate or garrisoned), and I didn’t get the other bonuses. Neither in real (at Castle in the beginning of next week), nor as part of the info message (as shown in your screenshot). What game version are you using? And can you check if it’s only the R-click msg, or if the bonus applies for real (probably yes, but who knows…).


I have nothing against this. Just tell me how to use it. Registered on sf.net as ivan91.
Thanks for remarks. I’ll try to make cleaner code next time:)

I’d wish to make some more changes in castle handling. Should I replace “old version” if I’ll find it?


I’m using H3 Complete + WoG 3.58f with the script updates and whatnot. It did give the advertised amount of creatures at the start of week. I tested it in SoD too and it was the same there.


Then I’m really puzzled (I have the same, except for the “whatnot”). :stuck_out_tongue:

Can you post here a saved game? If then I can reproduce it, perhaps it’s not related to the game version but sth else.


Here’s a save of a non-WoGified SoD map in WoG, and the test map itself.
Statue of Legion test.zip (142 KB)


Okay, this is even more strange: after trying your save, I am not able anymore to reproduce the behavior I had before, even with my own saves (or I was really blind to something before…). I’ll try to think if there’s anything I’m still missing, that led to the behavior I had before, but for now I’ll take it that the correct H3 behavior is the one you mention, namely that the bonus of the artifacts forming the Statue of Legion apply as well, when the hero wearing it is in town. I’ve updated #318 accordingly already.

Thanks. :wink:


Not only you :confused:
I’ll restore that behavior with next patch (tomorrow i think)


I’m working on properly implementing combination artifacts, and I’m thinking it might just be better if these artifacts just set their own bonus and when the engine retrieves the bonus for an artifact it will collect the bonuses of the constituents at that point.
This means if a future script were to change a constituent artifact the combination artifact is automatically updated, same would also apply to adding/removing constituents.


I gave you SVN access, so you can commit your changes directly to the repository. “Howtos” depends on your system - I’m using TortoiseSVN but it’s win-only. You may look for graphic Linux-compatible tools (like RapidSVN) or commit changes via terminal. Syntax is described in SVN manual ( svnbook.red-bean.com/en/1.1/re06.html ).
But I believe that if you’ve managed to download sources from SVN, you shouldn’t have troubles with commiting your changes :wink:

I gave you also developer status at bugtracker, so you are now able to mark bugs as resolved.

You may, especially when modifying that parts of interface. The code works same, it’s an aesthetic difference.

There are also other new features introduced to GUI btw. It’s possible to create parent - children hierarchy between CIntObject-s. If you put in the constructor OBJ_CONSTRUCTION_CAPTURING_AL macro, all CIntObjects subsequently created (till the end of scope) will be automatically assigned as children of that object. It’s useful for automatic propagation of some calls among object.

All activate/deactivate/show/delete/move* calls will be also made on children controls if the defActions (for parent) and recActions (for child) are appropriately set.
Child object position is initialized with it’s parent position (if it parent->defActions & SHARE_POS), so it’s easy to code relative positions between GUI objects (just add rel pos to base value).

For examples of usage of modern GUI features, you may check CAdventureOptions, CSlider, or pregame classes.

Makes sense. :slight_smile:
It will be great to have combination artifacts working, good luck on your work!
Have you thought about spell scrolls handling?
I don’t have a clean idea for them. They all have the same ID and therefore are considered as same artifacts. But we need another number to store id of spell - not possible within current CArtifact class. :confused:


Actually command to download source code is present in linux.readme - but not diffs & commit commands. I’ve find out how to work with svn already so this is not a problem now.

On weekend I’ve made a small bits of kingdom overview window. Nice place to try out them :wink:

Tow, after your recent changes (rev 1402 & 1403) I can not compile VCMI.

Connection.cpp:50: error: template<class Serializer> class CISer used without template parameters
Connection.cpp:51: error: template<class Serializer> class COSer used without template parameters

Files from previous revision compiled fine. But some new glitches appeared after end turn. Is that problem only on my side?


@spell scrolls
I think its best to make spellscrolls make new type of herobonus tempspell and spellID willl be subtype of bonus

or is it indifrientable now which scroll contains which spell?

also maybe add to CArtifact a field
vector BindHBonuses
which will be handled as adding those bonuses to hero when equip?

then it will be able not only to check type of spellscroll, but also thee will be a possibility to enchant artifact which may be good subject of future mods


It seems inevitable that there needs to be a new class representing artifact instances which can hold these parameters. Will be a lot of work migrating from using CArtifact to that, though. :frowning:

@majaczek: The problem is that the CArtifact class represents the different artifact types, not each individual artifact you might encounter in the game. If you were to change bonuses of a CArtifact you would change bonuses for all artifacts of that type.


So do class CScroll which derives from CArtifact :unamused:

OR maybe create class which have function I though - class of individual artifact with CArtifact as friend… so things like Spell scrolls will be resolved also my thought of enchanting etc.


There’s also a third error on line 52, which is some sort of circular dependency that gcc cannot figure out how to deal with (and neither can I).

Connection.h: In instantiation of 'CPointerLoader<CISer<CConnection>, CGTownInstance>': 
Connection.h:482:   instantiated from 'void CISer<Serializer>::registerType(const T*) [with T = CGHeroInstance, Serializer = CConnection]'
../lib/RegisterTypes.cpp:23:   instantiated from 'void registerTypes1(Serializer&) [with Serializer = CISer<CConnection>]'
../lib/RegisterTypes.cpp:158:   instantiated from 'void registerTypes(Serializer&) [with Serializer = CISer<CConnection>]'
Connection.cpp:52:   instantiated from here
Connection.h:437: warning: 'class CPointerLoader<CISer<CConnection>, CGHeroInstance>' has virtual functions but non-virtual destructor

which keeps repeating until gcc throws in the towel.

Deriving CArtifact won’t really change anything. If you stick to just using CArtifact you’ll end up with a solution where you assign new artifact IDs to each spell, giving lots of different spell scroll artifacts.


understood. of course if CArtifact is assigned via artifact ID it wont change anything :unamused: I feel like silly boy

I vote for Class which represent artifact instances! so instead having ID only we will be having this class. the most important field of this class will be artifact ID (so acces to CArtifact isn’t broken) and a few more fields like
vector UniqueArtHeroBonuses;
which will be applied if that instance of artifact will be worn and disapplied if it will be worn out. so we will make only one CArtifact for scrolls like now, we will have difffrient scrolls, and we will be able to enchant artifacts (add bonuses to one instance keeping rest the same) for a mod


The code you pasted is not about an error but about a warning.
But actually G CC is right. CBasicPointerLoader / *Saver should receive virtual destructors. And be deleted. There was a possible leak.
Fixed in r1411.
Are there any further issues with code on G CC?

PS. Mysterious g cc issue hit again :confused:


Looks like I was wrong, I thought I fixed the errors and then it still wouldn’t compile. It compiles now (r1412), but the compile time is a bit extreme. gcc didn’t get stuck as I thought, it just took literally 3 minutes to compile CGameState.cpp. I suppose it’s a pathological template case. Here’s a slew of warning output for these files, lots of non-virtual destructor complaints.
All in all it takes about half an hour to compile VCMI from scratch with gcc :open_mouth:, but I guess that’s to be expected with C++.

I also still get the gcc posting issue, but I use a zero-width workaround by typing