#3637 closed defect
GOB: not compilable under ps2 toolchain in MINGW/MSYS
| Reported by: | SF/mac_es78 | Owned by: | sunmax |
|---|---|---|---|
| Priority: | normal | Component: | Port: PS2 |
| Version: | Keywords: | ||
| Cc: | Game: |
Description
Hello,
I was trying to compile ScummVM 0.11.1 whith ps2 toolchain under MINGW/MSYS. I got an error when I tried to include gob engine.
The 'offending' code is in engines/gob/map_v1.cpp and was introduced in revision 28591.
The error is: Assertion failure in tc_gen_reloc at ../../gas/config/tc-mips.c line 1440
The real code is in the init() method:
if (_passMap || _itemsMap) return;
It seems that you can't use the pointers like that. For example, if (_passMap[0] || _itemsMap[0]) will work...
My ps2 toolchain is based in: binutils-2.14, gcc-3.2.2, newlib-1.10.0 and an updated ps2sdk.
All other engines in the branch are compilable.
Thanks
Ticket imported from: #1902303. Ticket imported from: bugs/3637.
Change History (14)
comment:1 by , 16 years ago
comment:2 by , 16 years ago
Yes, it is a compiler bug... No, it doesn't work either, the same error. I had tried it already, forgot to mention in the initial comment.
comment:3 by , 16 years ago
Ok, trying to create a proper code to bypass this bug. I found that the problem are not the pointers, the problem is the assignament of the internal variable whith the result of the or by the compiler.
So, this code compiles:
bool initialized = _passMap || _itemsMap; if (initialized) return;
May I create a patch, or it is not necessary?
comment:4 by , 16 years ago
| Owner: | set to |
|---|
comment:6 by , 16 years ago
Hi there Max!
Am I allowed to modify GOB code to make it MINGW-friendly?
Should I put an #ifdef for the PS2+MINGW case?
Please let me know, -max
comment:7 by , 16 years ago
It seems there is no need for #ifdef -- apparently some simple changes help to work around this nasty compiler bugs. Does any of these work? if (_passMap != NULL || _itemsMap != NULL) return; or if (_passMap) return; if (_itemsMap) return;
If so, change it accordingly, and add a comment pointing to this bug, stating something like "Do not modify this check unless you have to, else it may break certain cross compilers (see bug #XYZ)".
Anyway, I really wish those tool chains wouldn't use buggy compilers from the stoneage... I mean, even gcc 3.2.*3* would be better than 3.2.2; not to mention 3.3.6 or 3.4.6 (I am not even going to bother thinking about 4.x ;).
comment:8 by , 16 years ago
| Status: | new → pending |
|---|
comment:9 by , 16 years ago
This tracker item is pending response by the submitter; we cannot continue processing it before that happens. As a consequence, its status has been set to "Pending". It will automatically revert to "Open" once a new comment is made to this item. If no response is made within 14 days, it will automatically be closed.
Thank you.
comment:10 by , 16 years ago
Hi,
I'm mac_es78 but whith the correct username :) Sorry for the delay.
None of them works. The code I wrote in my third comment works however:
bool initialized = _passMap || _itemsMap; if (initialized) return;
comment:11 by , 16 years ago
Ops! Forgot to comment, I have tried it against 0.12.0 branch as the trunk for ps2 is uncompilable because of the recent savefiles changes
comment:12 by , 16 years ago
| Status: | pending → closed |
|---|
comment:13 by , 16 years ago
This Tracker item was closed automatically by the system. It was previously set to a Pending status, and the original submitter did not respond within 14 days (the time period specified by the administrator of this Tracker).
comment:14 by , 5 years ago
| Component: | → Port: PS2 |
|---|
So this is a compiler bug. Does it work if you use if (_passMap != 0 || _itemsMap != 0) return; instead?