Astrée is a static code analyzer that proves the absence of runtime errors and invalid concurrent behavior in safety-critical software written or generated in C.
Astrée primarily targets embedded applications as found in aeronautics, earth transportation, medical instrumentation, nuclear energy, and space flight. Nevertheless, it can just as well be used to analyze any structured C programs, handwritten or generated, with complex memory usages, dynamic memory allocation, and recursion.
Astrée is sound — that is, if no errors are signaled, the absence of errors has been proved.
Astrée analyzes whether the C programming language is used correctly and whether there can be any runtime errors during any execution in any environment. This covers any use of C that, according to the C99 standard, has undefined behavior or violates hardware-specific aspects.
Additionally, Astrée reports invalid concurrent behavior, violations of user-specified programming guidelines, and various program properties relevant for functional safety.
Astrée detects any:
TerminateTaskon a task with unreleased resources),
Astrée is sound for floating-point computations
and handles them precisely and safely. All possible rounding errors, and their
cumulative effects, are taken into account. The same is true
for −∞, +∞ and
NaN values and their effects
through arithmetic calculations and comparisons.
The seamlessly integrated RuleChecker lets you check your code for compliance with MISRA, CWE, ISO/IEC, and SEI CERT C coding rules. You can easily toggle individual rules and even specific aspects of certain rules. The tool can also check for various code metrics such as comment density or cyclomatic complexity. Custom extensions for your own in-house coding guidelines are available on request.
Using RuleChecker in conjunction with the sound semantic analyses offered by Astrée guarantees zero false negatives and minimizes false positives on semantical rules. No standalone MISRA checker can offer this, and no testing environment can match the full data and path coverage provided by the static analysis.
In 2003, Astrée proved the absence of any runtime errors in the primary flight-control software of an Airbus model. The system’s 132,000 lines of C code were analyzed completely automatically in only 80 minutes on a 2.8GHz 32-bit PC using 300MB of memory (and in only 50 minutes on an AMD Athlon 64 using 580MB of memory). Since then, Airbus France has been using Astrée in the development of safety-critical software for various plane series, including the A380.
In 2008, Astrée proved the absence of any runtime errors in a C version of the automatic docking software of the Jules Verne Automated Transfer Vehicle (ATV), enabling ESA to transport payloads to the International Space Station. This was the first fully automatic docking maneuver not performed by a Russian vessel.
The global automotive supplier Helbako in Germany is using Astrée to guarantee that no runtime errors can occur in their electronic control software and to demonstrate MISRA compliance of the code.
Bosch Automotive Steering replaced their legacy tools with Astrée and RuleChecker, resulting in significant savings thanks to faster analyses, higher accuracy, and optimized licensing and support costs.
AREVA employs Astrée for verification of their safety-critical TELEPERM XS platform that is used, among other things, for engineering, testing, commissioning, operating and troubleshooting nuclear reactors.
MTU Friedrichshafen uses Astrée to demonstrate the correctness of control software for emergency power generators in power plants. Together with its qualification package, Astrée is part of the IEC 60880 certification process.
A world leader in motors and ventilators for air-conditioning and refrigeration systems, ebm-papst is using Astrée for fully automatic continuous verification of safety-critical interrupt-driven control software for commutating high-efficiency EC motors for ventilator systems.
Astrée offers powerful annotation mechanisms for supplying external knowledge and fine-tuning the analysis precision for individual loops or data structures. Detailed messages and an intuitive GUI guide you to the exact cause of each potential runtime error. Actual errors can then be fixed, and in case of a false alarm, the analyzer can be tuned to avoid it. This allows for analyses with very few or even zero false alarms.
The analyzer can also run in batch mode for easy integration into established tool-chains.
Your usage of Astrée can be qualified according to DO-178B/C, ISO 26262, IEC 61508, EN-50128, the FDA Principles of Software Validation, and other safety standards. We offer special Qualification Support Kits that simplify and automate the qualification process.
Most static analyzers do not consider all possible runtime errors. Others specifically focus on the most probable ones.
As a result, almost all competing tools can only be used for static testing (i.e. finding some frequently occurring bugs), but never for verification (i.e. proving the absence of any runtime errors).
In contrast, Astrée is sound. It always exhaustively considers all possible runtime errors. It will never omit pointing out a potential runtime error. This is crucial for verification of safety-critical software. At the same time, Astrée is capable of producing exactly zero false alarms.
Certain static analyzers (e.g. those relying on theorem provers) require programs to be annotated with lots of inductive invariants.
Astrée usually requires very few annotations. It can even run completely automatically on certain types of programs, without relying on any help from the user.
Many analyzers cannot be scripted at all. Others can, but won’t let you access their analysis results outside of their proprietary viewer. This actively prevents you from automating the analysis, e.g. as part of your nightly build process.
In contrast, Astrée offers you complete access to the analysis engine in batch mode, and lets you freely export the analysis results and further process them in any way you choose, no strings attached.
Many static analyzers have high computational costs (typically several hours of computation per 10,000 lines of code). Others terminate out of memory, or may not terminate at all.
In contrast, Astrée is efficient and easily scales up to real-world programs in industrial practice.
As an example, in order to analyze actual flight-control software with 132,000 lines of C code, even on a slow 2.8GHz PC Astrée takes a mere 80 minutes. Faster machines will get you faster results. Multicore parallel or distributed computation is supported.
General-purpose static analyzers aim at analyzing any application written in a given programming language. They can rely on language related properties to find potential runtime errors. Specialized static analyzers put additional restrictions on the applications so as to be able to take specific program structures into account.
In contrast, Astrée is domain-aware. It thus knows facts about application domains that are indispensable to make sophisticated proofs. For example, for control/command programs, Astrée takes the logic and functional properties of control/command theory into account.
In static program analysis, there is always a trade-off between analysis precision and analysis cost. Analyzers that are precise are usually also very slow, while fast analyzers usually lack precision.
In contrast, Astrée is parametric, allowing you to freely trade speed for precision and vice versa. The level of abstraction used for analysis can be easily tailored to your very own requirements.
Astrée consists of several modules (so-called abstract domains). These modules can be assembled and parameterized to build application-specific analyzers that are fully adapted to a particular application domain or specific end-user requirements.
In case of false alarms, Astrée can be easily extended by introducing additional modules that enhance the precision of the analysis.
General-purpose static analyzers usually suffer from low precision in terms of false alarms, i.e. spurious warnings about errors that can actually never occur at runtime. The ratio of false alarms to the number of basic C operations typically ranges between 10% and 20%.
Specialized analyzers achieve a better precision of 10% or less. However, even a very high selectivity rate of only 1 false alarm for every 100 operations is usually unacceptable for large real-world applications. For example, on a program with 100,000 operations, a selectivity rate of only 1% yields 1000 false alarms.
In contrast, thanks to its modularity and domain-awareness, Astrée can be made exceptionally precise, often to the point of producing exactly zero false alarms. This has been repeatedly proven in industrial practice, e.g. when analyzing primary flight-control software for Airbus.
Astrée not only builds upon decades of research in static program analysis, but also incorporates the latest ongoing research, always staying well ahead of any competition.
Request your free trial package today.