MartinNowack

@andrewvaughanj I'm glad this works. I had a look at yours and @kren1 examples. A couple of optimisation (e.g. initialisation of stack-local structs) do not generate debug information for memory...

@ccadar Thanks for the review and suggestion. One goal of the implementation was to keep as much symbolic information as possible. I was thinking about this, the downside from this...

Just as a short catch-up: The reason this happens is due the virtualised file pointers (https://github.com/klee/klee/blob/24badb5bf17ff586dc3f1856901f27210713b2ac/runtime/POSIX/fd.c#L197), the mapping is specific to each state but unfortunately is incorrectly propagated to the...

@Alkaid-Benetnash Good catch! This is indeed an issue, thanks for reporting.

As we don't support LLVM 2.9 anymore, we can use `cl::getRegisteredOptions()` to implement this (https://llvm.org/docs/CommandLine.html#the-cl-getregisteredoptions-function). This should be straight forward now.

(Let's do that before this issue celebrates its 5th-year anniversary. 🎉 )

@yxliang01 The idea of the current state of the Docker image is to be able to use KLEE and to modify its code base. It's kind of having a minimal...

@yxliang01 I do agree, a more lightweight image would be nice. Due to the different systems, versions, configs, the build system should simplify many things. But I do agree it's...

@lzaoral Unfortunately yes. It's part of a bigger upgrade process.

According to `TODO.txt`, as we currently use github most for communication, I'm afraid a file will just get outdated to easily. For the "possible improvements", I would rather like to...