[LLVMbugs] [Bug 12152] New: some piped commands hang infinitely when input pipe is broken/terminated

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=12152

             Bug #: 12152
           Summary: some piped commands hang infinitely when input pipe is
                    broken/terminated
           Product: new-bugs
           Version: trunk
          Platform: Macintosh
        OS/Version: MacOS X
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: fang at csl.cornell.edu
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified


gtimeout 5m  /Users/fang/local/src/LLVM-svn/gcc40-cmake-build/bin/./clang -cc1
-internal-isystem
/Volumes/Isolde/builds/LLVM/gcc40-cmake-build/bin/../lib/clang/3.1/include
/Users/fang/local/src/LLVM-svn/clang/test/CodeGen/2006-01-13-StackSave.c
-emit-llvm -o - | opt -std-compile-opts |     llvm-dis | grep llvm.stacksave

When the built clang times out (reason not yet known), the opt program seems to
hang on an infinite loop.
Perhaps there's a missing check for input stream state somewhere, and it's
trying to read from a closed pipe infinitely?
Looks like culprit of several tests is hanging 'opt'.

Such failures plague TONS of CodeGen tests on powerpc-darwin8, which is now
seeing a lot more timeouts on clang than before, indicating possible infinite
looping.

possible workaround might be to attempt to 'timeout' every piped command, but
that requires hacking more than the lit dispatcher test scripts.

These tests were last known to run to completion with r150570, current version
tested is r151632.

To reproduce, trying running a test with sufficiently long input piping into
opt, and kill the input pipe, or time it out.

-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.