#4579 closed defect (fixed)
T7G: Compiling in Visual C++ 2008 Express causes error
| Reported by: | SF/aroenai | Owned by: | bluegr |
|---|---|---|---|
| Priority: | normal | Component: | Engine: Groovie |
| Version: | Keywords: | build | |
| Cc: | Game: | 7th Guest |
Description
Compiling ScummVM in MS Visual C++ 2008 Express Edition causes a "Trying to read a script byte at address 0xF102, while the script is just 0x4113 bytes long!" error when starting The 7th Guest. The game works normally in the daily build, I keep getting this error on both my computers after I compile the project.
This is a fresh svn checkout of r43847.
Ticket imported from: #2847539. Ticket imported from: bugs/4579.
Attachments (1)
Change History (16)
comment:1 by , 15 years ago
| Priority: | normal → high |
|---|
comment:2 by , 15 years ago
comment:3 by , 15 years ago
| Priority: | high → normal |
|---|
comment:4 by , 15 years ago
Have you tried a full build after doing a clean? If "starting" means straight after loading from the launcher, then it almost appears an endian issue since there is a jump to 0x02F1, but there is no reason for the swap to occur.
Several other devs also use MSVC (myself included) and haven't encountered this issue.
comment:5 by , 15 years ago
Yes, it does the same thing whether I do a fresh svn checkout or use the clean option from the build menu. Right, it's from the launcher, I double click the name and that error pops up.
comment:6 by , 15 years ago
Actually, I've just encountered this for the first time now. For some reason, it's being triggered in Release mode while I usually run in Debug mode which is a regression _somewhere_ (I'm 95% sure I've run in Release before fine), but I don't know where, or where to start looking :/
comment:7 by , 15 years ago
I can confirm that it works normally when you set it to debug, release mode is what causes it.
I'm not sure, but I think it happened sometime around the time when the output directory for the final exe changed.
comment:8 by , 15 years ago
I was right, here's where it started:
r42286 - Groovie works fine in release
r42287 - (doesn't compile, preparing for r42288 I guess)
r42288 - "Trying to read a script byte at address 0xF102, while the script is just 0x4113 bytes long!"
by , 15 years ago
| Attachment: | vs2008grooviefix.patch added |
|---|
makes groovie load when built in release mode
comment:9 by , 15 years ago
maybe groovie.vcproj is getting the wrong options? I just added some stuff that was different in the previous file version and after I compiled in release mode it worked, so there's probably a lot of junk that isn't actually needed in my patch but maybe it's a good starting point.
comment:10 by , 15 years ago
Thanks for tracking it down. I'll try to take a look over the next few days.
comment:12 by , 15 years ago
I got caught up with RL for a while, but I did get a chance to take a look at it with [md5] on IRC 2 days ago. It appears we can "fix" the problem by changing the optimization level in the Release vsprops file from 1 to 3, however rather than just committing this, it would be nice to work out why changing the optimisation level fixes it.
aroenai: Can you confirm that just the change in optimisation level removes the error?
comment:13 by , 15 years ago
I changed the optimization settings to the default (i.e. full optimization), and this seems to have fixed it
Closing this
comment:14 by , 15 years ago
| Owner: | set to |
|---|---|
| Resolution: | → fixed |
| Status: | new → closed |
comment:15 by , 15 years ago
After some discussion on IRC, we found an ambiguity in the Groovie multi-byte script variable reads which should certainly fix any future issues.
Please don't change the priority of tickets.