KYRA: LOL: Cannot Use Vaelan's Cube On Scottia's Barrier

[alois-kende]

When rightclicking in front of the barrier, it says I am too close to cast a spell. If I take a step back, the cube does do its visual effect but if has no effect on the barrier.

CD version
Android 5.1.1

[alois-kende]

save file

[digitall]

Tested with latest git master on Linux x86_64 with the attached savegame. You need to choose Vaelan's cube from the inventory and right click on one of the characters while in front of the barrier. The first time does an animation and the second time removes the barrier and the cube is used up.

The upshot is that I can't replicate this issue with the current git master.

@alois-kende: What version of ScummVM on Android are you using? The version is below the ScummVM logo on the launcher screen.

[digitall]

Possibly outdated... but pending feedback from user as can not replicate on desktop.

[alois-kende]

I am using v2.0.0 of scummVM on Android. No update seem to be available on the play store.
Android v5.1.1 for reminder.

I had issues with the various shortcuts for right click (not any would work) so I mapped a physical key as the right click to counter this issue. I have no idea if this can induce the behavior I see on my device. Apart from that, I can think of nothing special about my device/setup.

[alois-kende]

I just tested again. It does work. Maybe I didn't have v2.0.0 a month ago when submitting the ticket, which is odd...

[alois-kende]

I reacted a bit to quickly. It does not work all the time. It only worked once and then by reloading the save the problem reappears. So I guess there is still a problem.

[alois-kende]

I think I found how to reproduce the problem and how to avoid it.

The only way for the cube to be working on my device is :

• Reboot the phone
• Start scummVM
• Start Lands of Lire
• Load the save from WITHIN the game

-> Using cube on barrier works.

Now if I leave scummVM and don't reboot, it fails whatever I try.
If after rebooting, I start scummVM and load the save from OUTSIDE the game with the scummVM load save option, it fails as well.

Hope this helps ! (And if someone found this by googling their problem, they have a workaround :) )

[digitall]

Hmm... Now I can replicate from both ScummVM launcher load and the in-game loading with ScummVM 2.1.0git3935-g442c578223 (Nov 22 2018 17:19:49) on Linux x86_64.

Looks like an unstable bug. Will run valgrind and see if I can locate any issue.

[digitall]

Ah found the cause:

==21543== Invalid read of size 2
==21543==    at 0xD7EC4C: Kyra::LoLEngine::castSpell(int, int, int) (lol.cpp:1993)
==21543==    by 0xDA909E: Kyra::LoLEngine::olol_castSpell(Kyra::EMCState*) (script_lol.cpp:2077)
==21543==    by 0xDB1F41: Common::Functor1Mem<Kyra::EMCState*, int, Kyra::LoLEngine>::operator()(Kyra::EMCState*) const (func.h:451)
==21543==    by 0xD1472F: Kyra::EMCInterpreter::op_sysCall(Kyra::EMCState*) (script.cpp:306)
==21543==    by 0xD140DA: Kyra::EMCInterpreter::run(Kyra::EMCState*) (script.cpp:208)
==21543==    by 0xE90CC8: Kyra::LoLEngine::runItemScript(int, short, int, int, int) (items_lol.cpp:311)
==21543==    by 0xE86BA1: Kyra::LoLEngine::clickedPortraitEtcRight(Kyra::Button*) (gui_lol.cpp:1178)
==21543==    by 0xDCF035: Common::Functor1Mem<Kyra::Button*, int, Kyra::LoLEngine>::operator()(Kyra::Button*) const (func.h:451)
==21543==    by 0xE8B2D6: Kyra::GUI_LoL::processButtonList(Kyra::Button*, unsigned short, signed char) (gui_lol.cpp:2168)
==21543==    by 0xE687A8: Kyra::KyraEngine_v1::checkInput(Kyra::Button*, bool, int) (kyra_v1.cpp:343)
==21543==    by 0xE84B4C: Kyra::LoLEngine::gui_updateInput() (gui_lol.cpp:768)
==21543==    by 0xD7A292: Kyra::LoLEngine::runLoop() (lol.cpp:892)
==21543==  Address 0x1e044cda is 6 bytes before a block of size 63 alloc'd
==21543==    at 0x4C30F25: calloc (vg_replace_malloc.c:711)
==21543==    by 0x2013DF8: Graphics::Surface::create(unsigned short, unsigned short, Graphics::PixelFormat const&) (surface.cpp:74)
==21543==    by 0x200027E: Graphics::TTFFont::cacheGlyph(Graphics::TTFFont::Glyph&, unsigned int) const (ttf.cpp:597)
==21543==    by 0x1FFF160: Graphics::TTFFont::load(Common::SeekableReadStream&, int, Graphics::TTFSizeMode, unsigned int, Graphics::TTFRenderMode, unsigned int const*) (ttf.cpp:254)
==21543==    by 0x2000585: Graphics::loadTTFFont(Common::SeekableReadStream&, int, Graphics::TTFSizeMode, unsigned int, Graphics::TTFRenderMode, unsigned int const*) (ttf.cpp:661)
==21543==    by 0x1EE0D9F: GUI::ThemeEngine::loadScalableFont(Common::String const&, Common::String const&, int, Common::String&) (ThemeEngine.cpp:1482)
==21543==    by 0x1EE1193: GUI::ThemeEngine::loadFont(Common::String const&, Common::String const&, Common::String const&, int, bool) (ThemeEngine.cpp:1540)
==21543==    by 0x1EDD043: GUI::ThemeEngine::addFont(GUI::TextData, Common::String const&, Common::String const&, int) (ThemeEngine.cpp:526)
==21543==    by 0x1EEB45F: GUI::ThemeParser::parserCallback_font(Common::XMLParser::ParserNode*) (ThemeParser.cpp:187)
==21543==    by 0x1EF23CA: GUI::ThemeParser::CustomXMLKeyLayout::doCallback(Common::XMLParser*, Common::XMLParser::ParserNode*) (ThemeParser.h:53)
==21543==    by 0x1EF2344: GUI::ThemeParser::keyCallback(Common::XMLParser::ParserNode*) (ThemeParser.h:53)
==21543==    by 0x213B8BF: Common::XMLParser::parseActiveKey(bool) (xmlparser.cpp:196)
==21543== 

[athrxx <athrxx@…>]

In 4b2dc67c:

KYRA: (LOL) - fix #10764 (Cannot Use Vaelan's Cube)

[athrxx]

I have made a commit to the git repository that should fix the bug.

It seems that we didn't import all of the static data from the original executable. That caused the invalid memory access and the random results for the vaelans cube.

Can you actually test this, although you play on Android?