Builds, tests passed, jenkins tests are ok, feature looks good, I do not see more problems in code
- Queries
- All Stories
- Search
- Advanced Search
- Transactions
- Transaction Logs
Advanced Search
Sep 19 2019
builds in debug and without pch release
From Debug build
I see two cases when to add header to precompiled.
- It is included in a lot of files ( what means more than 15 with small files)
- It is big file in terms of code and other includes and is used more than 5 times.
Sep 18 2019
looks good to me
Sep 17 2019
try to fail after timeout
add timeout
fix robocopy
fix robocopy
Sep 16 2019
broke test.exe build with --jenkins-tests and so jenkins
Sep 15 2019
no, but might be interesting improvement
add flags for gloox and collada
split testing so release tests follow right after build of release version so in case they do not pass, there will be no build of debug version
Tests passed
file: source\ps\CStr.cpp line 195 : [C-style cast]: return (long)Pos; line 206 : [C-style cast]: return (long)Pos; line 217 : [C-style cast]: return (long)Pos; line 232 : [C-style cast]: return (long)Pos; line 242 : [a++ -> ++a]: for (size_t i = 0; i < length(); i++) line 243 : [C-style cast]: NewString[i] = (tchar)_totlower((*this)[i]); line 251 : [a++ -> ++a]: for (size_t i = 0; i < length(); i++) line 252 : [C-style cast]: NewString[i] = (tchar)_totupper((*this)[i]); line 349 : [a++ -> ++a]: for (size_t i = 0; i < length(); i++) line 365 : [C-style cast]: ss << "\\u" << std::hex << std::setfill('0') << std::setw(4) << (int)(unsigned char)ch; line 381 : [a++ -> ++a]: for (Left = 0; Left < length(); Left++) line 396 : [a++ -> ++a]: for (Left = 0; Left < length(); Left++) line 447 : [C-style cast]: return (size_t)fnv_hash(data(), length()*sizeof(value_type)); line 448 : Comments should start with capital: // janwas 2005-03-18: now use 32-bit version; 64 is slower and line 461 : [a++ -> ++a]: for (i = 0; i < len; i++) line 483 : [C-style cast]: *(str++) = (tchar)native; line 503 : [C-style cast]: Serialize_int_4(buffer, (u32)len); line 505 : [a++ -> ++a]: for (i = 0; i < len; i++) file: source\ps\tests\test_CStr.h line 82 : a+ -> a +: u8* buf = new u8[len+1]; line 158 : [WARNING]: File is missing include guard
Sep 14 2019
Sep 13 2019
In D2287#95144, @Vulcan wrote:Successful build - Chance fights ever on the side of the prudent.
Link to build: https://jenkins.wildfiregames.com/job/vs2015-differential/154/display/redirect
less code
address comment by wraitii so this fixes also scenario where unit changes ownership and technologies change Health values based on player
Do we want only one type of units in a battalion?
yes, that is one of the basic idea of batalions, but there would be allowed special units like banner carrier, hero, commander.
Sep 12 2019
fixed in rP22892
static cast to ensure u16 result
Sep 11 2019
Sep 10 2019
Sep 9 2019
add entry to template, fix repair command back
yes possibly but first it would be better just have this one as it is a lot faster than computing squares and this just changes unitai code in minimal scale
Sep 8 2019
yes icons should have own folder
Sep 6 2019
@Stan
units have different sound for order_repair and order_build at least in templates
issue is that i unit_actions.js is only command repair and so always is used and played order_repair even if target is just foundation and not damaged building
Sep 3 2019
Aug 27 2019
Aug 17 2019
Aug 16 2019
builds without-pch
Aug 15 2019
Aug 9 2019
Jul 28 2019
Fixed in rP22567
@Stan can you please check if it works for you?
Jul 27 2019
just for record there is already similar logic:
Hero Elephant = Hero or Elephant
Hero+Elephant = Hero and Elephant
!Hero = no Hero
Jul 26 2019
In rP22511#35467, @elexis wrote:In rP22511#35454, @Angen wrote:This introduced Access violation reading from ticket: https://trac.wildfiregames.com/ticket/5539 not fixed by later patch.
Sure it was this commit? As I understand it's the placement new in rP22545 that is not supported by VS2013?
Jul 25 2019
Reason for this issue with VS2013 is that memset was removed and looks like vs13 does not do default initialisation of that values.
If vs13 is only compiler having that problem, then we should cut off vs13 build already with D1396 since this is making vs13 builds not functional at all.
Grid.h(234): Assertion failed: "0 <= i && i < m_W && 0 <= j && j < m_H"
This introduced Access violation reading from ticket: https://trac.wildfiregames.com/ticket/5539 not fixed by later patch.
Jul 24 2019
Fixed in rP22544
Build on windows is green, no crashes related to glsl or postprocessing, sound plays in atlas as should.
My RAM usage for replay from https://trac.wildfiregames.com/ticket/5530
Jul 23 2019
I think we should never fail in case some mod has wrong data anyway, unless it is fail of whole mod io.
After that, I try to come with some not downlandable mod line for mods with wrong data.
Also do not forget \public\gui\reference\viewer\viewer.js for right click window (encyclopedia or whatever is it name)
In D2114#88408, @elexis wrote:Would it be cleaner to only push valid elements rather than adding all elements and removing the invalid?
I think it would.
In D2114#88408, @elexis wrote:Are there reasons to inform the users that a mod was not signed?
At first: Users should not be able to download not signed mods or mods with outdated signature.
And they cannot do much about it, so spamming them with the messages about all unsigned mods would not be kind.
In D2114#88408, @elexis wrote:Is it considered a mistake if such an element appears, and if it is, would it be reasonable to inform the user so as to have it reported to the mod signers?
If you mean that there is no information about signature, what means there is no object, it is error, because that metadata contains information also about dependencies.
But in the first place ingame downloander permits users to downloand not - signed mods, because that means that mod is not trusted.
also remove test assert as we ignore mod silently
Builds and tests are green on windows.
I do not fully agree with restricted classes for splash damage. As was said restricted classes and preferred classes are something for unitai and petra to choose better target and do not waist attack at something it should not attack.
If there will be in future real time detection for projectiles collision, the fact if that projectile or splash damage, death damage will deal some damage to the given entity should be based on damage type vs armour type and not restricted classes. But this is just my personal opinion. :)
if we allow left and right how would one cancel preselected action?
Jul 22 2019
selection details that part i posted is when you hover over that sword and shield when selected entity, it takes data from entity state so uses cmpGarrisonHolder
@Nescio
in selection_details
Engine.GetGUIObjectByName("attackAndArmorStats").tooltip = [ getAttackTooltip, getSplashDamageTooltip, getHealerTooltip, getArmorTooltip, getGatherTooltip, getSpeedTooltip, getGarrisonTooltip, getProjectilesTooltip, getConsumerTooltip, getPayedTooltip, getResourceTrickleTooltip, getLootTooltip ].map(func => func(entState)).filter(tip => tip).join("\n");
vs
selection_panels
let template = GetTemplateData(data.item); let tooltips = [ getEntityNamesFormatted, getVisibleEntityClassesFormatted, getAurasTooltip, getEntityTooltip, getEntityCostTooltip, getGarrisonTooltip, getPopulationBonusTooltip, showTemplateViewerOnRightClickTooltip ].map(func => func(template));
it displays chosen tooltip and formulate it based on data it gets. It does not gets data by itslef, data are given to it.