[llvm-bugs] [Bug 27903] New: Seg fault due to miscompile from new stack coloring data flow analysis

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=27903

            Bug ID: 27903
           Summary: Seg fault due to miscompile from new stack coloring
                    data flow analysis
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: tejohnson at google.com
                CC: llvm-bugs at lists.llvm.org, thanm at google.com
    Classification: Unclassified

During a clang bootstrap with ThinLTO on linux using the gold linker,
llvm-tblgen started seg faulting. I bisected this to r270559 - Rework/enhance
stack coloring data flow analysis. It also goes away when compiling with
-no-stack-coloring.

I was able to narrow this down to utils/TableGen/RegisterInfoEmitter.cpp, and
further to when _ZN4llvm11raw_ostreamlsEl is imported into
RegisterInfoEmitter.cpp from lib/Support/raw_ostream.cpp. This is eventually
inlined into
_ZN12_GLOBAL__N_119RegisterInfoEmitter9runMCDescERN4llvm11raw_ostreamERNS1_13CodeGenTargetERNS1_14CodeGenRegBankE. 

To reproduce, untar the attached tblget.tgz. Change the paths in all.sh,
link.sh and run.sh to point to your own build and source trees. Then run all.sh
and it will run through 3 llc/link/run cycles. The first should fail and the
latter 2 should work. The all.sh script describes what is different about the
three. I have the llvm IR for the good and bad files in the same location as
the good and bad files but with the .ll extension.

I tried using bugpoint to narrow further, but couldn't get the compile steps to
link correctly with the rest of the native .o files.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20160527/02a55712/attachment.html>