Otherwise they will overwrite each other when we invoke the same
executable mutliple files. This was producing race conditions for a
device connected through NFS for me.
Note: This does not affect the Makefile based test-suite as it never
runs an executable more than once.
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@259907 91177308-0d34-0410-b5e6-96231b3b80d8
Note: This does not affect the Makefile based test-suite in any way as
the RunSafely -d option is only used by the
External/SPEC/**/CMakeLists.txt files.
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@259906 91177308-0d34-0410-b5e6-96231b3b80d8
This will be necessary for the upcoming SPEC support for cmake/lit which will
use the reference data coming with SPEC instead of the custom one in
the Externals directory.
This commit also reverses the setting for the RunSafely output format to
produce the traditional output format by default because that is still
used by most benchmarks. That way fewer lit.local.cfg files are
necessary.
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@255874 91177308-0d34-0410-b5e6-96231b3b80d8
This is useful for external benchmark suites which bring reference files
for stdout without the llvm-test-suite convention of having an exit code
behind the stdout output.
As discussed in D14678 I am going for post-commit review as cmake/lit
test-suite support is still in early development.
Differential Revision: http://reviews.llvm.org/D14679
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@254835 91177308-0d34-0410-b5e6-96231b3b80d8
Simply leaving out -l let's ssh and rsh pick a sensible default user
(and allows to specify them in .ssh/config for example).
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@254131 91177308-0d34-0410-b5e6-96231b3b80d8
bash's ulimit -f is in multiples of 1024 so previous the limit was 10GB not
10MB. Turns out when I made it actually be 10MB, that is too small for two
current tests.
- Set the limit to be 100MB instead.
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@153025 91177308-0d34-0410-b5e6-96231b3b80d8
- Simplifies a bunch more goop, we now are down to basically executing only two
main commands to run tests (the monitor process and the actual target process).
- Also kill off the random core file handling stuff, this isn't particularly useful.
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@153022 91177308-0d34-0410-b5e6-96231b3b80d8
This is not the right fix. TimedExec.sh needs to be replaced by adding
a timeout option to the timeit tool. But, this has been sitting in my
copy of the test-suite for months and I would like to be able to tell
others to use the option. Disabling TimedExec.sh is the only way to get
meaningful run times for short benchmarks.
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@151237 91177308-0d34-0410-b5e6-96231b3b80d8
and exit statuses that indicate that the test program was terminated
by a signal as failures.
Also, report failures by writing a string to the output file which
will trigger a diff when it is compared with output files from other
runs of the test.
And, having reported any detected errors in a way that will be
visible to the test system, always exit "successfully". RunSafely.sh's
own exit status is not used to determine if a test has passed or failed.
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@58154 91177308-0d34-0410-b5e6-96231b3b80d8
the program and the time command is properly separated. This used to work only
on shells that have a builtin "time" command (such as bash). This change makes
things work consistently among different shells (tested on bash, dash and
FreeBSD sh) and should be POSIX compatible.
A similar change is probably needed for the remote execution part, but someone
who is using that should have a look at it.
git-svn-id: https://llvm.org/svn/llvm-project/test-suite/trunk@51369 91177308-0d34-0410-b5e6-96231b3b80d8