fix #2 (bad type of some parameters for some z80/68k low-level commands) #86
Conversation
Bug introduced in the refactor in the 57733d5 commit
adriweb
changed the title
| @@ -540,7 +540,10 @@ int ti82_send_asm_exec(CalcHandle* handle, VarEntry * var) | |||
| endptr = (var->name[0] == 0x24 ? fpBase : tempMem); | |||
| buffer[8] = 0x2a; buffer[9] = LSB(endptr); buffer[10] = MSB(endptr); | |||
| /* ld de, -program_size */ | |||
| #pragma warning(push) | |||
There was a problem hiding this comment.
#pragma isn't portable in the first place, but would #pragma msvc ... be possible at least for the warning(disable:4146) ?
There was a problem hiding this comment.
Well I know, but gcc and clang just warn that it's an unknown pragma and ignore it anyway :P
But contrary to those two compilers with a compiler-name-as-first-arg, it looks like #pragma msvc isn't a thing...
So I suppose we could add #ifdef _MSC_VER/#endif for each. but that's whole lot of extra lines just to disable a warning, bleh.
There was a problem hiding this comment.
Taking a step back... fixing the code would be better :)
offset should be int32_t, and after adding an early bounds check against < 2 and >= 32768 (both models have 32 KB of physical RAM) on var->size below VALIDATE_VARENTRY(var), something like offset = -(int32_t)((uint16_t)var->size - 2); should do the job, hopefully without the compilers complaining about numerical value truncation or overflows.
There was a problem hiding this comment.
Yeah that's probably better. We'll have to compare the actual output in buffer just in case...
|
I'll close this PR since e5ad813 was integrated. |
And also add some safety flags to the cmake build...