|
From: Paul F. <pa...@fr...> - 2011-03-09 17:33:18
|
Hi I'm getting this error: ==9395== Use of uninitialised value of size 8 ==9395== at 0x5445E28: array_dyn_resize (vassign.c:1368) ==9395== by 0x1E92D537: ??? ==9395== by 0x1E830057: ??? ==9395== by 0x5A7A2E7F: ??? ==9395== by 0x1E7F030F: ??? ==9395== Uninitialised value was created by a stack allocation ==9395== at 0x4AEC3B0: exec_process (in [snipped].so) ==9395== vex amd64->IR: unhandled instruction bytes: 0xC0 0x36 0x7D 0x5A 0x0 0x0 ==9395== valgrind: Unrecognised instruction at address 0x5a7a2e88. This is in an app that assembles and executes code, so it could be a genuine error. That said, there are a lot of errors before this reaching this one. Does this look like a valid opcode? On RHEL53 64bit. A+ Paul |
|
From: John R. <jr...@bi...> - 2011-03-09 18:09:44
|
> vex amd64->IR: unhandled instruction bytes: 0xC0 0x36 0x7D 0x5A 0x0 0x0
The instruction bytes "C0 36" are officially undefined. The second byte
("modR/M" in the Intel nomenclature) 0x36 designates case 6==(0x7 & (0x36 >> 3))
which would be an 8-bit shift (SHL) with immediate constant shift count,
except that SHL and SAL are equivalent, so the case 4 is used for both,
and case 6 is undefined. The code generator forgot this special tweak.
--
|
|
From: Julian S. <js...@ac...> - 2011-03-10 11:46:55
|
> This is in an app that assembles and executes code, so it could be a > genuine error. Apps like that tend to die with very strange errors unless you run with --smc-check=all. Did you do that? J |