[llvm-commits] [llvm] r170350 - /llvm/trunk/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp

[Jim Grosbach]

On Dec 17, 2012, at 1:09 PM, "Malea, Daniel" <daniel.malea at intel.com> wrote:

> Actually, the MCJIT unit tests had their own memory manager before there was a suitable default. Now that there is one, they just use SectionMemoryManager, however, I think it would be easy to extend that class to allow for a custom validating memorymanager to be used  (see MCJITTestBase.h:54).
> 
> For the most part, the MCJIT unit tests are target-independent in that an IRBuilder is used to construct some very basic global variable and function definitions that are run through MCJIT and executed.
> 
> One thing that I should mention that Darwin (and Cygwin) are currently in the list of operating systems for which the tests are skipped, so if you are developing on one of those platforms, adjust the list on MCJITTestBase.h:77.
> 

The tests used to pass, IIRC, for 64-bit Darwin. It's just i386 where they still have some problems. If someone can confirm that and commit a change to enable 64-bit OSX testing for the MCJIT, it would be much appreciated.

-j

> I'd be happy to review any MCJIT unit test improvements!
> 
> 
> Best of luck,
> Dan
> 
> On 2012-12-17, at 3:53 PM, Chandler Carruth wrote:
> 
> Do you know of any RuntimeDyld tests that aren't target-specific? In
> principle we could have a unittest that loaded a provided object and
> checked the sections created (with a custom memorymanager?). There
> doesn't appear to be much precedent for that approach though.
> 
> Sadly, no... =/
> 
> Daniel Malea <daniel.malea at intel.com<mailto:daniel.malea at intel.com>>
> Intel Waterloo
> Phone: 519-772-2566
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits