UnitSync :: GetModOptionCount Crash

[danil_kalina]

Execute GetModOptionCount, crash in game The Cursed and Evolution RTS

try to execute in separate project, also crash

[bibim]

Same problem I have with SPADS on Windows I guess...

[danil_kalina]

I guess the same issue

[abma]

mantised: http://springrts.com/mantis/view.php?id=2916

[danil_kalina]

unitsync.log is empty
how to turn it on ?

[abma]

setting

LogSections = ArchiveScanner, VFS

in springsettings.cfg should enable some logging.

can you be more specific what functions you are calling?

this example doesn't crash: http://pastebin.com/R7izPY59

(written in python using the unitsync / ctypes wrapper from upq)

[danil_kalina]

we call GetModOptionCount()

games that are 100% crash: The Cursed 1.16, Evoultion RTS 1.9
OS: Windows

we'll try to turn the logging

[a1983]

UnitSync crashed exactly after getModOptionCount in The Cursed 1.16.
By the way, can I change LOG_LEVEL_INFO?

[hoijui]

abma was asking for what functions you are calling.
if you would really be calling only this one function, it would be no wonder that you crash, and i am sure you are not doing that.

i guess we can be somewhat sure that it is a problem that happens on windows only. that could mean it is related to a file-system call, or it has to do with access to non-explicitly initialized memory, or... maybe something else. :D

[a1983]

That's UnitSync Functions calls sequence that cause crash. Help me!!!

[abma]

sorry, i don't get a crash here, when i try to call these functions on windows...


can you please run the program with an attached debugger and get a stacktrace?

[Forboding Angel]

Abma, spring must be in isolation mode in order to get the crash

[abma]

it also doesn't crash for me in isolation-mode... :-/

please do a "bt full" with "gdb"...

[a1983]

Thx for responce :)
It's really weird crash. I can't catch exception by gdb, stack is not available. May be it's special to windows7 64 or mingw 4.4.0?

[hoijui]

you need debug symbols for untisync. they can be found here for current latest release:
http://springrts.com/dl/buildbot/defaul ... ync_dbg.7z
extract this archive, and make sure that gdb uses the file. i think you do it with the -s symbol-file cmd-line switch (or some gdb console command).

[a1983]

It's too hard to me :) I'm not strong in gdb debuging. But I got trace at last.
It looks like exception raised by "throw content_error()" in Option.cpp raise another uncaught exception. Or something else... )))

[hoijui]

nice!
thanks, and well done!

it looks like this archive contains bad options:
28f1e3632fb713d232c8c75f2373f2bf.sdp
this is a an archive managed by rapid, and as much as i know, you should be able to delete the file by hand, or .. maybe try to do a rapid cleanup first, and see if that removes it already.

edit:
to det/Licho/anyone else who knows... how can we find out the rapid tag that belongs to this archive, so we can test ourselfs?
thanks to jk, i found out that it is thecursed:version:1.16, will now extract mod options...

[a1983]

So why it cause crash? In code content_error created and crash is caused exactly by throw. It's strange, because I think exception must be caught and try to log error.
Of cause I can delete this file but where I can get good mod file. And what to do with broken archive?

P.S. I downloaded that archive just now using rapid_gui (The Cursed 1.16)

[hoijui]

(i edited my last post)

modoptions.lua from thecursed:version:1.16

---- KING OF THE HILL OPTIONS ----
{
key = 'koth',
name = 'King of the Hill Settings',
desc = 'Settings for King of the Hill mode.',
type = 'section',
},
{
key="KOTH",
name="King Of The Hill",
desc="The one with ...",
type="number",
section= 'koth',
min = 0,
max = 10080,
step = 1,
def = 0,
},

indeed we should catch that error and log it in unitsync (it is a bug).
also the cursed should be fixed of course.

to uninstall:
rapid unpin thecursed:version:1.16

[a1983]

You are pretty fast