Tag Archives: csmith

Why do signed overflows so often do what programmers expect?
Pascal Cuoq on 29 March 2012

Semi-serious musings During the Frama-C random testing experiment described at length on this blog and this page we found a few bugs in Csmith too. John Regehr one of the Csmith developers and not entirely coincidentally a co-author of the article linked in the previous post is also a co-author...

Read More

Csmith testing again
Pascal Cuoq on 16 January 2012

My presentation Friday at the U3CAT meeting was on the topic of Frama-C Csmith testing. Several posts in this blog already describe facets of this work (it has its own tag). Yet another angle can be found in this short article draft. Said draft by the way will soon need...

Read More

Explaining why Csmith matters even more than previously anticipated
Pascal Cuoq on 4 December 2011

Csmith as a static analyzer fuzzer A new version of Csmith, the generator of random defined C programs, was released a few days ago. This is the version that many functions in Frama-C Nitrogen were debugged against. Conversely a few bugs in the development versions of Csmith characterized by programs...

Read More

How to waste a Friday evening
Pascal Cuoq on 21 October 2011

Here is a quick recipe for completely wasting a Friday evening: use Csmith to generate a program that GCC happens to mis-compile in 32-bit into an executable that produces the same result as the correct 64-bit compilation of the same program; mess up the Clang double-check by forgetting the -m32...

Read More

New alarm category
Pascal Cuoq on 26 September 2011

There won't be many changes in the value analysis' documentation in Nitrogen. For lack of time, the new options, some of which were alluded to in this blog, will remain in their \to be documented" state. But documenting them more fully can always be done here too once they are...

Read More