[llvm-commits] [LNT] r168813 - /lnt/trunk/lnt/tests/compile.py
Michael Gottesman
mgottesman at apple.com
Wed Nov 28 13:17:03 PST 2012
Author: mgottesman
Date: Wed Nov 28 15:17:03 2012
New Revision: 168813
URL: http://llvm.org/viewvc/llvm-project?rev=168813&view=rev
Log:
[compile tests] Previously, I was just allowing the codesize_util to
return -1 if it was a failure. This distorts the lnt graph by screwing
up the scale. Now I changed the utility to return "fail" instead,
which I can detect and emit a failure.
Modified:
lnt/trunk/lnt/tests/compile.py
Modified: lnt/trunk/lnt/tests/compile.py
URL: http://llvm.org/viewvc/llvm-project/lnt/trunk/lnt/tests/compile.py?rev=168813&r1=168812&r2=168813&view=diff
==============================================================================
--- lnt/trunk/lnt/tests/compile.py (original)
+++ lnt/trunk/lnt/tests/compile.py Wed Nov 28 15:17:03 2012
@@ -463,16 +463,17 @@
cmd = codesize_util + [os.path.join(build_base,
binary_path % format_args)]
result = subprocess.check_output(cmd)
- bytes = long(result)
- success = True
-
- # For now, the way the software is set up things are going to get
- # confused if we don't report the same number of samples as reported
- # for other variables. So we just report the size N times.
- #
- # FIXME: We should resolve this, eventually.
- for i in range(variables.get('run_count')):
- samples.append(bytes)
+ if result != "fail":
+ bytes = long(result)
+ success = True
+
+ # For now, the way the software is set up things are going to get
+ # confused if we don't report the same number of samples as reported
+ # for other variables. So we just report the size N times.
+ #
+ # FIXME: We should resolve this, eventually.
+ for i in range(variables.get('run_count')):
+ samples.append(bytes)
except OSError as e:
if e.errno != errno.ENOENT:
raise
More information about the llvm-commits
mailing list