On 03/04/2015 07:08 AM, Jose Fonseca wrote:
On 04/03/15 12:58, Jose Fonseca wrote:
On 04/03/15 04:20, Matt Turner wrote:
On Tue, Mar 3, 2015 at 8:07 PM, Brian Paul <bri...@vmware.com> wrote:
---
  src/gallium/auxiliary/gallivm/lp_bld_misc.cpp | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
b/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
index 5210acc..e2578cf 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
+++ b/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
@@ -493,7 +493,7 @@
lp_build_create_jit_compiler_for_module(LLVMExecutionEngineRef *OutJIT,
     builder.setMCPU(MCPU);
  #endif

-   ShaderMemoryManager *MM;
+   ShaderMemoryManager *MM = NULL;
     if (useMCJIT) {
  #if HAVE_LLVM > 0x0303
         BaseMemoryManager* JMM =
reinterpret_cast<BaseMemoryManager*>(CMM);
--
1.9.1

I'm guessing it's the 'delete MM;' that generates the warning? If so,
shouldn't you just wrap it in #if HAVE_LLVM > ...?

I think MCJIT is the only option in llvm >= 3.6? The code could
probably be trivially refactored to avoid the assert(0) as well.

 From what I can tell, it seems like adding initializers is nearly
always the wrong thing to do.
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.freedesktop.org_mailman_listinfo_mesa-2Ddev&d=AwIGaQ&c=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEs&r=zfmBZnnVGHeYde45pMKNnVyzeaZbdIqVLprmZCM2zzE&m=eFrgqffTZg-fVfsgBecewE76PU07pVGCehKDjd2rc2Q&s=NmdJuPH44mfwP-iaquREc8ZPTboVxNFpMQhNXlMv2FY&e=




Yes, this is probably related to
https://urldefense.proofpoint.com/v2/url?u=https-3A__bugs.freedesktop.org_show-5Fbug.cgi-3Fid-3D89387&d=AwIGaQ&c=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEs&r=zfmBZnnVGHeYde45pMKNnVyzeaZbdIqVLprmZCM2zzE&m=QY0p_bDuP-f3LPn9v8UCnsJc9MFPxmK5yIA_AQLYfpk&s=QtRiKhT6OE4Vfm-MYI7OR5SNu_iqhBPIOUnOp_5tmTc&e=


But unfortunately I haven't found time to catch up with it.

On closer look, this is a different issue.  As Matt said, the issue is
the delete for older LLVM.

That said, `delete NULL` is safe, hence Brian's patch is good.

I think this is simplest. Matt, if you look at the rest of the neighboring code, MM gets assigned (basically) if (using MCJIT and llvm >= 303) or llvm < 306 so it's kind of complex condition. Just init'int to null is the simplest solution.

I'll take Jose's comment as an R-b.

-Brian


_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev

Reply via email to