[compiler-rt] [libFuzzer] Fix merge-posix.test file size test

[ndrewh]

This test uses ulimit -f 1 to test what libFuzzer does when trying to create a file > 1KB. However, the none of the input files used by this test are actually >= 1KB, so there's no reason to expect this test to pass.

This test appears to be passing on accident since the "control file" happens to be > 1KB, but this is not always the case depending upon the length of the path where the test is run from.

This modifies the test to ensure that one of the input file is actually >1KB.

[llvmbot]

@llvm/pr-subscribers-compiler-rt-sanitizer

Author: Andrew Haberlandt (ndrewh)

Changes

This test uses ulimit -f 1 to test what libFuzzer does when trying to create a file > 1KB. However, the none of the input files used by this test are actually >= 1KB, so there's no reason to expect this test to pass.

This test appears to be passing on accident since the "control file" happens to be > 1KB, but this is not always the case depending upon the length of the path where the test is run from.

This modifies the test to ensure that one of the input file is actually >1KB.

---

Full diff: https://github.com/llvm/llvm-project/pull/168137.diff

1 Files Affected:

• (modified) compiler-rt/test/fuzzer/merge-posix.test (+1-1)

diff --git a/compiler-rt/test/fuzzer/merge-posix.test b/compiler-rt/test/fuzzer/merge-posix.test
index 2721668fb9706..edd91324e5a92 100644
--- a/compiler-rt/test/fuzzer/merge-posix.test
+++ b/compiler-rt/test/fuzzer/merge-posix.test
@@ -14,7 +14,7 @@ RUN: echo ....U. > %tmp/T2/2
 RUN: echo ...Z.. > %tmp/T2/3
 RUN: echo ...Z.. > %tmp/T2/4
 RUN: echo ....E. > %tmp/T2/5
-RUN: echo .....R > %tmp/T2/6
+RUN: (echo .....R; for i in {1..1024}; do echo -n X; done) > %tmp/T2/6
 
 # Check that we can report an error if file size exceeded
 RUN: (ulimit -f 1; not %run %t-FullCoverageSetTest -merge=1 %tmp/T1 %tmp/T2 2>&1 | FileCheck %s --check-prefix=SIGXFSZ)

[DanBlackwell]

@cmtice I remember you were doing some work regarding using the 'internal shell' (#160728); does that work with the syntax used in this change?

[ndrewh]

fwiw, before I made this change I did look and found other tests doing similar shell scripting

llvm-project/compiler-rt/test/asan/TestCases/Darwin/duplicate_os_log_reports.cpp

Line 13 in 49d5bb0

// RUN: for I in {1..3}; do \

I'm definitely not sure that this is the best / most portable way to write out a 1024B file 😄. We could just call into python3 (other tests do that), or use dd (but i didn't see any other tests doing that).

[ndrewh]

This is another option:

RUN: %python -c "f = open('%tmp/T2/6', 'wb'); f.write(b'.....R'.ljust(1025, b'X')); f.close()"

[DanBlackwell]

Internal shell supports redirects, so this would do I believe:

RUN: %python -c "print('X' * 1024, end='')" > %tmp/T2/6

[DanBlackwell]

Update looks good, thanks!

[boomanaiden154]

I remember you were doing some work regarding using the 'internal shell' (#160728); does that work with the syntax used in this change?

Yes, it will.