DEBUG: Code Questions (Compiling, Bugs, Revision Reports, Etc)

Ekmek · April 10, 2007, 16:25

I'm creating this thread because the revision reports just reports changes and the playtest thread really should only

be for experiences (good, bad, crashes).

So this thread will be for problems arising from revision changes that break compiling or bugs detected. I know the

code now has a lot of TODOs in there and I don't have a way of tracking them all down but I'll try to list them here.

CURRENT ISSUES, BUGS, REQUESTS

* Barbarian build lists
* Player[m_owner] = Null
* Invisible Units "revealed"
* Terraform Resouce Bug
* No Piracy Agreements
* AI Tile Imp placement/choice
* Barbs only attack cities
* Barbs ZOC not updating
* Missing savegame graph
* AI sea Invasion Lag
* Brazil scenario persistence
* Turnlength not implemented
* Map Import
* Growth rate
* Vanishing Space planes
* Ingame Menu - New Game Crash
* Empire Manager commerce miscalculates

FIXED ISSUES
* Worker Bug && here
* Unit Upgrades only graphic not DB traits
* Scenario Loading
* Great Library errors
* Unit Herald Rework
* drawTransitionTile memory crash

Ekmek · April 10, 2007, 16:28

Originally posted by E
Since revision 709

this line of code (used in my sectarianhappiness but may cause crashes elsewhere) compiles but causes a crash:

Code:

		if(m_cityStyle != g_player[m_owner]->GetCivilisation()->GetCityStyle()) {

why is that?

Originally posted by Martin Gühmann
Sinply if g_player[m_owner] is NULL then it crashs.

By the way E you shouldn't post such questions here, this thread is for revisions reports, there are a lot of other threads you can use.

-Martin

putting a

Code:

if (g_player[m_owner] != NULL) {

doesn't work either. But this code used to work before 709 so I think a MSVC6++ compatibility was broken. Fortunately this crash can be avoided by having the userprofile for sectarianhappiness to be off. But this line is the same as cultureonly units and other stuff throughout the code so something is broken.

Ekmek · April 10, 2007, 16:29

am I the only one that gets a lot of assert errors when trying to run civctp_dbg.exe? at start up it hits me with setnew.h ln 52 and then a game i get diplomat.cpp ln870
then it may throw up other diplomat errors in line 206,207,208 or 543

Martin Gühmann · April 10, 2007, 16:31

Well, you could always use the debugger, you can set breakpoints everywhere within the code.

And by the way E, I thought you were a mod now, so you can move posts into new threads. Or are there some technical difficulties?

-Martin

Ekmek · April 10, 2007, 17:01

I'm limited in my mod powers... but i did delete some of the stuff from the revision reports (the last twoposts) for clean up sake.

I would use the debugger but that is almost unusable now because of the errors i get in setnew.h and diplomat.cpp

Martin Gühmann · April 10, 2007, 17:10

Originally posted by E
I would use the debugger but that is almost unusable now because of the errors i get in setnew.h and diplomat.cpp

You don't need to run a turn with the AI, as you said the code in question is called when you open the city manager.

-Martin

Ekmek · April 10, 2007, 17:37

it happens on start up. You don't get this problem?

Code:

You don't need to run a turn with the AI

are you saying there is a way to shut off the AI?

Martin Gühmann · April 10, 2007, 17:41

Originally posted by E
it happens on start up. You don't get this problem?

I get it, too. But as far as I know it is just one assert. So that isn't aproblem.

Originally posted by E

Code:

You don't need to run a turn with the AI

are you saying there is a way to shut off the AI?

No, I don't say to turn of the AI. But since your code is called as you said when you open the city manager, you don't need to test it on the AI turn. So just open the city manager and don't press the end turn button. And by the way the scenario editor can help you, too.

-Martin

cap601 · April 11, 2007, 16:03

Having upgraded to v.719 I decided to have a look at the scenarios. Attempting to run any scenario (excl. World Map and AE Mod) gives a series of messages like this one referencing a variety of databases. Attempting to go continue past all of them works but on trying to reenter the main window then this appears. After some random clicking further parts of a menu appear but nothing actually does anything. The game does close via Alt+F4.

Ekmek · April 11, 2007, 16:41

Originally posted by cap601
Having upgraded to v.719 I decided to have a look at the scenarios. Attempting to run any scenario (excl. World Map and AE Mod) gives a series of messages like this one referencing a variety of databases.

thanks cap, i'll look at it. the good its referencing is located in the AE mod which is weird because the scenarios should begoing to the default folder not he AE mod. This might be a case where you still have the AE_mod "loaded" and then try to start a scenario. I guess to work around it is to only start a scenario on start-up not to click through the various ones. That is, if you haven't already tried that

Attempting to go continue past all of them works but on trying to reenter the main window then this appears. After some random clicking further parts of a menu appear but nothing actually does anything. The game does close via Alt+F4.

I've found that going past them never works. I do know that there is always a slic error on start-up with alex scenario. I've wanted to redoit by resaving the game as a scenario but Fromafar said its good to keep it as is (not exactly sure why)

cap601 · April 11, 2007, 17:21

Originally posted by E
thanks cap, i'll look at it. the good its referencing is located in the AE mod which is weird because the scenarios should begoing to the default folder not he AE mod. This might be a case where you still have the AE_mod "loaded" and then try to start a scenario. I guess to work around it is to only start a scenario on start-up not to click through the various ones. That is, if you haven't already tried that

The errors do appear even if the AE mod is not loaded before the scenario.

cap601 · April 12, 2007, 15:39

I can see it's buggy but how is unit upgrading meant to work? I haven't run across any options to do it at any point yet.

Ekmek · April 12, 2007, 16:04

unit upgrading is suupposed to work under a few conditions:

1) your unit is in a city
2) the technology is available for a new unit (see in the unit.txt UpgradeTo line)
3) you have the gold to do it (gold rush modifier times the difference in shield value)
4) you 'sentry' the unit in a city.

i think thats all the details. the method was based on the unitupdater code in SAP

for the record it used to work good to the playterster eye but it was messy in the code. Martin worked to fix that but I think he left off the change the actual unit part (not it just changes the sprite)

Ekmek · April 14, 2007, 01:17

I think this is the source of the worker bug

the current code is first with the '*/' stuff and i put an older version of code in. I have to test it though

Code:

void CityData::ChangePopulation(sint32 delta)
{
/*   //this code was change and maybe source of worker bug since workers done twice
	m_population += delta;
	m_numSpecialists[POP_WORKER] += (sint16)delta; //this might be extra
	
	if (delta < 0) 
    {
		m_numSpecialists[POP_WORKER] += (sint16)delta;
		if(m_numSpecialists[POP_WORKER] < 0) {
			sint32 needToRemove = -m_numSpecialists[POP_WORKER];
			m_numSpecialists[POP_WORKER] = 0;

			for (sint32 i = 0; i < POP_MAX && needToRemove > 0; i++) 
            {
				if(m_numSpecialists[i] >= (sint16)needToRemove) {
					m_numSpecialists[i] -= (sint16)needToRemove;
					needToRemove = 0;
				} else if(m_numSpecialists[i] >= 0) {
					needToRemove -= m_numSpecialists[i];
					m_numSpecialists[i] = 0;
				}
			}
		}
	}
*/

//original code (from rev 632 archive)
	m_population += delta;
	

	if(delta > 0) {
		
		m_numSpecialists[POP_WORKER] += (sint16)delta;
	} else {
		
		
		m_numSpecialists[POP_WORKER] += (sint16)delta;
		if(m_numSpecialists[POP_WORKER] < 0) {
			sint32 needToRemove = -m_numSpecialists[POP_WORKER];
			m_numSpecialists[POP_WORKER] = 0;

			sint32 i;
			for(i = 0; i < POP_MAX && needToRemove > 0; i++) {
				if(m_numSpecialists[i] >= (sint16)needToRemove) {
					m_numSpecialists[i] -= (sint16)needToRemove;
					needToRemove = 0;
				} else if(m_numSpecialists[i] >= 0) {
					needToRemove -= m_numSpecialists[i];
					m_numSpecialists[i] = 0;
				}
			}
		}
	}
	AdjustSizeIndices();
	UpdateSprite();

	if(g_network.IsHost()) {
		g_network.Block(m_owner);
		g_network.Enqueue(m_home_city.AccessData(), this);
		g_network.Unblock(m_owner);
	}

	if(m_population <= 0) {
		g_gevManager->AddEvent(GEV_INSERT_AfterCurrent, GEV_KillCity,
		                       GEA_City, m_home_city.m_id,
		                       GEA_Int, CAUSE_REMOVE_CITY_UNKNOWN,
		                       GEA_Player, -1,
		                       GEA_End);
	}
}

cap601 · April 14, 2007, 08:57

Wouldn't that be fixed if you set the marked line in an else statement at the end (i.e. it only runs with delta>=0)? That appears to be the only major change between the two code blocks (excl. part from AdjustSizeIndices(); onwards)

Fromafar · April 14, 2007, 16:47

Oops. I now remember being busy modifying this piece of code when getting interrupted by a RL event. I have completed the modification in revision 721, so it should be working again.

Fromafar · April 14, 2007, 17:02

Originally posted by E
putting a

Code:

if (g_player[m_owner] != NULL) {

doesn't work either. But this code used to work before 709 so I think a MSVC6++ compatibility was broken. Fortunately this crash can be avoided by having the userprofile for sectarianhappiness to be off. But this line is the same as cultureonly units and other stuff throughout the code so something is broken.

g_player[m_owner] being NULL would make g_player[m_owner]->GetCivilisation()->GetCityStyle() crash, but there are more possibilities.
Use the debugger to check if maybe

m_owner is "invalid" (e.g. PLAYER_UNASSIGNED)
g_player[m_owner] is OK, but GetCivilisation returns NULL

Fromafar · April 14, 2007, 17:22

Originally posted by E
am I the only one that gets a lot of assert errors when trying to run civctp_dbg.exe? at start up it hits me with setnew.h ln 52 and then a game i get diplomat.cpp ln870
then it may throw up other diplomat errors in line 206,207,208 or 543

The setnew.h assert is because of MSVC6 being too old to support the C++ standard properly.
You don't mention the circumstances for the diplomat asserts, but these could occur when disbanding in neutral territory. You prevented a crash with your update in revision 719 by not really using it, but you left in a reference to the diplomacy data of an unassigned player.

BTW: This thread is a mess, and my replies only seem make it worse

. I prefer your previous idea of having a separate (non-topped) thread per issue.

Martin Gühmann · April 14, 2007, 19:38

Originally posted by cap601
The errors do appear even if the AE mod is not loaded before the scenario.

And what kind of errors appear, for instance which error messages in particular? I tested it on my setup and the Alex Scenario loaded fines both on final and debug version. And no difference between revion 717, 719 and 721.

-Martin

cap601 · April 15, 2007, 05:57

I'm not going to do a full list as there are too many but the following databases have things missing:-
- Icon (e.g. ICON_GOOD_ARCTIC_MOUNTAIN_THREE)
- UnitBuildList (e.g. UNIT_BUILD_LIST_SPECIAL_SCIENTIST)
The messages are of this form

This is from loading the Alexander scenario immediately after running the game. Similar messages appear for the other scenarios but I believe the actual messages are different (IIRC one of them mentioned the Strategy data

Martin Gühmann · April 15, 2007, 16:08

Well, actually Fromafar did already some compatibility fixes which actually just restored loadability, but prevented the AI from usage the improved strategies.txt. Well obviously E never included those files. Furtuantely this makes it easer to archieve real compatibility. I updated UnitBuildList.txt and uniticon.txt so that the scenario can use the default strategies.txt. However there seems still to be a little incompatibility with the AI slider setting. Obviously city growth is slowed down in the Alex scenario that causes some problems since the AI tries to set the city growth at least to a rate of one pop per 30 turns. Well this can be fixed by modifying the strategies.txt. But maybe there is another solution.

Anyway the Nuclear Dentente scenario seems just to need an updated uniticon.txt and the Seven Smurai scenario an updated uniticon.txt as well and a modified goods.txt.

-Martin

cap601 · April 16, 2007, 12:48

Time for another bug:-
The Great Library entries for the different terrains show the entries for other terrains (example). There also appears to be (mucked up) entries for "bridge" and "special" which may be related.

Ekmek · April 18, 2007, 19:34

When I get a chance I'll use this thread to be the directory for current and fixed bugs. Slowly I'll migrate these to other threads in the meantime i have a readme to work on.

cap601 · April 19, 2007, 16:57

In addition to the previous bugs I've discovered a reference to a space colony. In this case the text is that of the last page visited before this one and so changes.

EDIT - I've done a bit of poking around with this one. The space colony reference can be removed by:-
-Commenting out the line #UNIT_SPACE_CITY "Space Colony" in gl_str.txt
-Commenting out the line #DESCRIPTION_UNIT_SPACE_CITY "junkstring_DESCRIPTION_UNIT_SPACE_CITY" in junk_str.txt
-Commenting out the whole of UNIT 57 in Units.txt

Not sure how this would affect other languages (gl_str and junk_str are in /English but Units is in /Default

How would I go about submitting this?

Also I'm going to look into fixing the others in a similar way.

EDIT2 - I believe I've found the true cause. There are a series of different definitions of objects with the property GLHidden. This appears to be broken as these all appear in the GL without a correct description. Here's a list of locations for these properties. Good luck in finding where in this code this happened.

Ekmek · April 19, 2007, 17:44

this might be a cde issue with the handling of the flag GLhidden.

if you want to submit stuff here is the source code repository
http://ctp2.darkdust.net/

but for in game text files you can upload them to apolyton and I can add them in. But hopefully martin will look into the errors IIRC he fixed this problem before

cap601 · April 19, 2007, 18:03

The text files depend on whether we want to prune out unneeded strings etc. Some should be easy to remove (e.g. Space colony) but some of the others are probably impossible (e.g. ORDER_MOVE). With a fix to GLHidden then it probably isn't worth it though it may be best for neatness sake.

cap601 · April 20, 2007, 05:42

I've prodded around in the code files and I believe the GLHidden problems are due to the file greatlibrary.cpp and specifically the function void GreatLibrary::UpdateList( DATABASE database ). In here it checks for the flag when clicking any of the side buttons and does not display the objects with the GLHidden flag (this works). However, when searching (case DATABASE_SEARCH) this is not done and the objects do appear on searches. Unfortunately I don't know where any check would go.

Hope that makes sense and helps.

Martin Gühmann · April 20, 2007, 07:38

Originally posted by cap601
In addition to the previous bugs I've discovered a reference to a space colony. In this case the text is that of the last page visited before this one and so changes.

EDIT - I've done a bit of poking around with this one. The space colony reference can be removed by:-
-Commenting out the line #UNIT_SPACE_CITY "Space Colony" in gl_str.txt
-Commenting out the line #DESCRIPTION_UNIT_SPACE_CITY "junkstring_DESCRIPTION_UNIT_SPACE_CITY" in junk_str.txt
-Commenting out the whole of UNIT 57 in Units.txt

Not sure how this would affect other languages (gl_str and junk_str are in /English but Units is in /Default

How would I go about submitting this?

Actually we don't want to change here anything for two reasons. First removing unit 57 from the unit databse breaks savegame compatibility and second even if we have other priorities we haven't given up to restore the space layer.

Originally posted by cap601
Some should be easy to remove (e.g. Space colony) but some of the others are probably impossible (e.g. ORDER_MOVE).

ORDER_MOVE is used that is the first one you can click on the order's button bank.

Originally posted by cap601
I've prodded around in the code files and I believe the GLHidden problems are due to the file greatlibrary.cpp and specifically the function void GreatLibrary::UpdateList( DATABASE database ). In here it checks for the flag when clicking any of the side buttons and does not display the objects with the GLHidden flag (this works). However, when searching (case DATABASE_SEARCH) this is not done and the objects do appear on searches. Unfortunately I don't know where any check would go.

That's indeed the problem.

-Martin

cap601 · April 25, 2007, 17:26

An old bug but if you save the game whilst a space plane is in space flight it vanishes. It still appears in the Unit Manager when you first load the file but vanishes on the next turn.

cap601 · April 26, 2007, 12:56

I've also noticed a few mistakes in the GL's text. A corrected copy is attached but I can't guarantee I've caught all of them. Any chance of adding it?