[llvm] r182085 - Minor changes to the MCJITTest unittests to use the correct API for finalizing

Kaylor, Andrew andrew.kaylor at intel.com
Fri May 17 09:58:01 PDT 2013


Hi Dave,

MIPS definitely does have that same weakness regarding multiple relocation passes.  I don't think you need to wait before disabling the test there too.

-Andy

-----Original Message-----
From: llvm-commits-bounces at cs.uiuc.edu [mailto:llvm-commits-bounces at cs.uiuc.edu] On Behalf Of David Tweed
Sent: Friday, May 17, 2013 5:58 AM
To: Amara Emerson
Cc: llvm-commits at cs.uiuc.edu
Subject: RE: [llvm] r182085 - Minor changes to the MCJITTest unittests to use the correct API for finalizing

Hi,

| The buildbot failure on the ARM pandaboard is due to an assertion failure:
| "Assertion `(*TargetPtr & 0x000F0FFF) == 0' failed." In the ARM
RuntimeDyld.

| This particular assertion is one I added last year as a sanity check 
| that the relocation target address didn't already have an addend in 
| it. Since
the
| object code may now be relocated before these are resolved, this
assumption
| may not be valid anymore.

| I don't removing this assertion should cause any problems.

Thanks for looking at this. (This slipped through because I tested on a -Asserts build to minimise the time the hardware was out of use; won't do that again...) I'm just running tests with that assertion removed, once that's confirmed I'll commit the assertion removal.

This patch has also caused a regression on the MIPS buildbot; I suspect the test is hitting MIPS equivalent of ARM's current inability to do multiple relocations, but I'll give it a couple of days for any MIPS expert to look at it before I disable the test on MIPS as well.

Cheers,
Dave





_______________________________________________
llvm-commits mailing list
llvm-commits at cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




More information about the llvm-commits mailing list