Imgur kills Odyssey? [SOLVED in 1.23]

Latest version. Generally stable.
However loading imgur now causes GR. Imgur used to work ok? I have made no changes to odyssey or MUI.

Crash log for task "OWB"
Generated by GrimReaper 53.16
Crash occured in module OWB at address 0x7F160E6C
Type of crash: DSI (Data Storage Interrupt) exception

Register dump:
GPR (General Purpose Registers):
0: 7F160DB4 5E38D920 00000000 5E38D92C 5CDF3B98 00000000 000000F0 219C2D3C
8: 5B037E18 CCCCCCCC 02212E08 02A7B396 0000015C 5E500740 00000000 5E4F8B08
16: 5E4FA3C8 5E4F8B64 DEAD0077 5E4FA3C8 5E7A0000 5E38DC78 5F368024 5E4F8B64
24: 5E38DC78 5F368024 00000052 5B60E0C0 5E38D930 5E38D92C 5E46FDFC 5B037E30


FPR (Floating Point Registers, NaN = Not a Number):
0: nan 0 0 -3.85298e+305
4: -3.85298e+305 -3.85298e+305 -4.4678e+307 -4.4678e+307
8: 1 255 1667 0.00252318
12: 4.5036e+15 1.39023e+09 0 0
16: 0 0 0 0
20: 0 0 0 0
24: 0 0 0 0
28: 0 0 1.39023e+09 1.39023e+09

FPSCR (Floating Point Status and Control Register): 0xA2002100


SPRs (Special Purpose Registers):
Machine State (msr) : 0x0200B030
Condition (cr) : 0x5BB25390
Instruction Pointer (ip) : 0x7F160E6C
Xtended Exception (xer) : 0x02028A78
Count (ctr) : 0x00000000
Link (lr) : 0x00000000
DSI Status (dsisr) : 0x020462B0
Data Address (dar) : 0x00000000



680x0 emulated registers:
DATA: 92B58200 00000000 00000000 00000000 00000000 00000000 00000000 00000000
ADDR: 6FFBAA00 92F9CF00 00000000 00000000 00000000 00000000 00000000 5E38C0D0
FPU0: 0 0 0 0
FPU4: 0 0 0 0



Symbol info:
Instruction pointer 0x7F160E6C belongs to module "OWB" (PowerPC)
Symbol: _ZN7WebCore16WebSocketChannel19didOpenSocketStreamEPNS_18SocketStreamHandleE + 0x140 in section 1 offset 0x0108FE48

Stack trace:
_ZN7WebCore16WebSocketChannel19didOpenSocketStreamEPNS_18SocketStreamHandleE()+0x140 (section 1 @ 0x108FE48)
_ZN7WebCore16WebSocketChannel19didOpenSocketStreamEPNS_18SocketStreamHandleE()+0x88 (section 1 @ 0x108FD90)
_ZN7WebCore5TimerINS_18SocketStreamHandleEE5firedEv()+0x30 (section 1 @ 0x110FC88)
_ZN7WebCore12ThreadTimers24sharedTimerFiredInternalEv()+0x140 (section 1 @ 0x17FD90)
_ZN7WebCore12ThreadTimers16sharedTimerFiredEv()+0x40 (section 1 @ 0x17FDF8)
_ZN7WebCore17fireTimerIfNeededEv()+0x38 (section 1 @ 0x16E9EC)
_ZN14WebViewPrivate21fireWebKitTimerEventsEv()+0x10 (section 1 @ 0xCB144)
_ZN7WebView21fireWebKitTimerEventsEv()+0x14 (section 1 @ 0xA8804)
_ZL28handleMM_OWBApp_WebKitEventsP6IClassPmP4_Msg()+0x90 (section 1 @ 0x44C4)
_ZL8dispatchP6IClassPmP4_Msg()+0x460 (section 1 @ 0xE740)
CustomClassDispatcher()+0xa0 (section 1 @ 0x41B4)
native kernel module intuition.library.kmod+0x0001824c
native kernel module intuition.library.kmod+0x00018470
native kernel module intuition.library.kmod+0x00008448
native kernel module intuition.library.kmod+0x00008088
_Z9main_loopv()+0x1e0 (section 1 @ 0x550)
main()+0xac (section 1 @ 0xFD4)
native kernel module newlib.library.kmod+0x000020a4
native kernel module newlib.library.kmod+0x00002d0c
native kernel module newlib.library.kmod+0x00002ee8
_start()+0x170 (section 1 @ 0x16C)
native kernel module dos.library.kmod+0x00024f18
native kernel module kernel+0x00042514
native kernel module kernel+0x00042594

PPC disassembly:
7f160e64: 809e8024 lwz r4,-32732(r30)
7f160e68: 7fa3eb78 mr r3,r29
*7f160e6c: 83890044 lwz r28,68(r9)
7f160e70: 48250c4d bl 0x7F3B1ABC
7f160e74: 7fe3fb78 mr r3,r31

System information:

CPU
Model: P.A. Semi PWRficient PA6T-1682M VB1
CPU speed: 1800 MHz
FSB speed: 900 MHz
Extensions: altivec

Machine
Machine name: AmigaOne X1000
Memory: 2097152 KB
Extensions: bus.pci bus.pcie

@ddni
Please, remove everything after "Expansion buses", it not need it , but make reading of thread harder.

As for crash itself, some sites may crash, make bug-reports on bugs.os4depot.net / odyssey.

EDIT: in case with inmgur.com site , its also crashes on 1.17 version of morphos version of odyssey, but works with 1.23, what mean, that once 1.23 for os4 will be released, that crash will gone (as many others with such stack trace).

In meantime you can do what kicko do : in url preferences disable java for sites which crash you.

Thanks kas1e.

The value 0xcccccccc in r9 clearly points out a duplicate Remove() of a node from an Exec list.

@Thore
From all my tests with 1.9 / 1.16 and now 1.23, i can say that all reports about sites crashes with such stack traces / disassemble / registers info its all webkit bugs, which they fix later. For example big bunch of crashing sites on 1.9, start to works on 1.16 (while some still crashes). And for sake of tests i do test morphos version as well, and the same sites the same crashes on the same versions, and stop crashes on recent ones.

@kas1e
QUOTE!
Quote:

@tlosm
I mean 1.23 on morphos, aos4 version still in process: currently i only port latest libcairo, libpixman, ffmpeg, sqlite, core of mediaplayer (with help from Daniel), compile odysseys's libjsc.a, libwtf.a and almost compile libwebcore.a (which is huge and big, and which have some moments to fix which wasn't need it in previous versions) , and once it all will be done, then there will be need to compile all that mui-amiga-native-code, which for sure will have something to fix for os4, something to rewrite which wasn't in previous version, as well as new class for autoform-filling is need it (calltips.mcc, but i hope Thore as usual will deal with it). After all will be compiles there for sure will be some linking problems after fix of which it will crashes on running for sure, then those fixes, then it will works but mediaplayer will be crashing for sure, also fixes and so on.

I.e. in progress, but slow notebook with cygwin take time for compiling as well as i not sit whole days over it too :)

Also will be good if someone will re-implement spellchecker.library , which is also in use by odyssey.

@kas1e

in any case is a great news :D

and only?... wow you make a great work in small time

@kas1e

Good good can say

Quote:

And this one is pretty intesting too!

@tlosm
Quote:
Yes, because all those parts was mostly ./configure; make with some very minor changes. All the other stuff going mostly like it was for previous versions, except mediaplayer and some new small problems related to compiler

@tboeckel

Quote:

Not really. I use memory allocators that also clear deallocated memory with this pattern. So it could be a double deletion in WebKit objects, most likely.

@kas1e

If you have an sqlite3 later than 3.7.5, can you please upload it to OS4Depot (when it's ready) along with source? Assuming it has an updated os_amiga.c. I have 3.7.5, but there are reasons I won't stick it on OS4Depot. I put the source on SpotUP's server (used for SdkServer. Go there manually via HTTP).

@Mick
My version of sqlite is pure ./configure; make:

Quote:

-DSQLITE_OMIT_WAL=1 need it for disable sys/mman.h dependence with all those MAP_SHARED, PROT_READ, etc.

--disable-dynamic-extensions need it for avoid using dynamic externsions (dlsym, dlclose, dlerror calls, etc)

I also have 3.7.5 version from HansJoerg which he use for TW, but currently i stick with my own to see if it will works at all, and if there will be any bugs (so if yes, then will reuse Hans's one).

@kas1e

Oh ok. I was under the impression that sqlite didn't work at all without the specific Amiga filesystem code (like trying to port findutils. If someone can do that, that'd be great). If in the future I port something that needs a later sqlite, then it's good to know I can just do a straight port.

@tboeckel and FAB.

“0xcccccccc” but how do you know, that this memory is never going to be allocated?

As I see it your creating potential problems, there most be better ways to deal whit duplicate Remove(), way not log the error some where else?

@LiveForIt

Quote:

If the memory is reallocated it should be reinitialised, thus no problem, if not this munging forces a "safe crash" and the developer can fix the code.

In the case of the list node, you *must not* rely on the ln_Pred and ln_Succ pointers being valid after the node is reomved, so there is no issue.

@ddni
Yep, it works on 1.23 aos4 port. Also notice that there was some ad-video block with video and speech : works too now ! I for first even didn't get what happens, some woman from that ad start to talk bull and i didn't realize at first that it from odyssey :)

@kas1e

Excellent! Thanks.
Great to hear of progress which I understand! The "libcurl etc etc" type discussions on the "official" Odyssey 1.23 thread go waaaay over my head.

From a users point of view, how close to a beta release are you?