[LLVMbugs] [Bug 376] NEW: llvm.py doesn't handle non-existent executables gracefully

bugzilla-daemon at cs.uiuc.edu bugzilla-daemon at cs.uiuc.edu
Sat Jun 19 11:36:14 PDT 2004


http://llvm.cs.uiuc.edu/bugs/show_bug.cgi?id=376

           Summary: llvm.py doesn't handle non-existent executables
                    gracefully
           Product: Test Suite
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: QMTest
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: rspencer at x10sys.com


While getting the optimized nightly tester ready, I ran into a problem where
QMTest and/or python and/or llvm.py would cause numerous "stuck" processes to
hang around on my system. This was caused by doing the following:

cvs co llvm
cd llvm
configure
gmake ENABLE_OPTIMIZED=1
cd test
gmake Feature.t

The problem is that the second gmake doesn't have ENABLE_OPTIMIZED=1 on the
command line so it attempts to execute the various tools (llvm-as, llc,
llvm-dis) from the Debug directory. The executables don't exist but the
ExecProgram python function in llvm.py doesn't clean up properly when it gets an
error like this. The forked process remains after a failed exec and those
processes accumulate quickly because the QMTest just keeps on going even though
the exit status from the failure should have been non-zero.

Although this is a usage problem, the fact that we have such an unreliable test
mechanism is troubling and we should really clean this up so that future
mistakes don't bring systems to their knees (I had to reboot my server to ensure
I didn't blow out the process table).



------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.




More information about the llvm-bugs mailing list