#2558 closed defect (fixed)
Problem with GP32 port.
| Reported by: | SF/wonst719 | Owned by: | SF/chrilith |
|---|---|---|---|
| Priority: | normal | Component: | Port: GP32 |
| Version: | Keywords: | ||
| Cc: | Game: |
Description
I tried to update port for the GP32 against latest SVN version (revision 21665). It compiles well, but it crashes when it tries to open config file useing hashed open. So I hacked into it, and...
config-manager.cpp Line 140: in ConfigManager::loadFile(...) file.cpp Line 206: in File::open(...) hashmap.h Line 322: in HashMap::contains(Key &) hashmap.h Line 302: in HashMap::lookup(Key &) func.h Line 31: in EqualTo::operator()(const String &, const String &) str.cpp Line 255: in String::operator ==(const String &) str.h Line 81: in String::c_str() CRASH
...the result is a hand-maded trace log. (I can't use GDB / debugger with GP32) I tried to put address checking in c_str() method. I usually get address 0x0c******, but this case I got 0xea000058. This prevents the GP32 port form working. I wonder if this happens because of compiler bug...?
Ticket imported from: #1466270. Ticket imported from: bugs/2558.
Change History (8)
comment:1 by , 18 years ago
comment:2 by , 18 years ago
* Oh. I made some mistake on the log. :-) c_str() was at line 71, not 81.
* The address I meant was "this" pointer. That position (0xea000058) is definitely not in the GP32 memory area.
* The string passed to the lookup() method and equal operator was "scummvm.ini".
* I can't get values what "_str", "_len" have... "this" already points to invalid position.
* I tried removing the config file. But the result was same.
Hope this answer helps...
comment:3 by , 18 years ago
| Owner: | set to |
|---|
comment:4 by , 18 years ago
Sounds as if this this might be the same problem that is troubling the PalmOS port (see <http://logs.scummvm.org/log.php?log=scummvm.log. 09Apr2006>, and search for PalmOS).
bascially, we are using two globals in common/file.cpp, which maybe don't get inited properly, namely static StringIntMap _defaultDirectories; static FilesMap _filesMap;
Chrilith, care to comment? How did you fix this?
If this is indeed the problem, we should change the code, to help you two guys. One possibility would be to change the two vars to be pointers, and then use "new" to allocate storage once they are accessed for the first time. Not elegant, but hopefully works.
Though I am a bit surprised, the old version of file.cpp contained StringList File::_defaultDirectories; and that seemed to work, at least I never saw complaints from either of you two... Hmm, there the variable is a static class member, does that make a difference, I wonder... ?
comment:5 by , 18 years ago
| Resolution: | → fixed |
|---|---|
| Status: | new → closed |
comment:7 by , 5 years ago
| Component: | --Unset-- → Ports |
|---|
comment:8 by , 5 years ago
| Component: | Ports → Port: GP32 |
|---|
I have no idea if this is due to a compiler bug, and obviously have no way to find out either :-).
Some questions: * You write "str.h Line 81: in String::c_str()" -> did you mean "71"? If not, your str.h differs from mine.
* When you say you checked the "address", about which address are you talking, precisely? The value of the "this" pointer? The value of "_str"?
* What values do "this", "_str", "_len" have upon the crash?
* If possible tell us the string passed to the lookup() method.
* If possible, tell us the string values of the two strings passed to the equal operator.
* Did you try removing your config file?
-> I am not sure if all this helps, it's all a bit like fumbling in the dark, but at least the info might help narrow down the issue.