@ccadar any chance for a review?

I am getting the following error while running klee.

KLEE: ERROR: Loading file ~/Desktop/usr/local/lib/klee/runtime/libkleeRuntimeFreestanding64_Debug+Asserts.bca failed: No such file or directory

I built klee using llvm-18 with uclibc, posix runtime support.

Yeah the CI is also failing on that, but I'm not sure how to investigate this since I did not change anything about the build scripts... Perhaps there is a silent error while compiling the runtime and it gets hidden from the build status?

@mrexodia thanks for your contribution. This looks find with me overall, although I was wondering why you replaced starts_with and end_with with find.

Other than that, our CI is failing and I would like to fix it first. Furthermore, we should add support for LLVM 17-19 to the CI, although for that we need to create some Docker images first, which @MartinNowack usually handles.

@mrexodia thanks for your contribution. This looks find with me overall, although I was wondering why you replaced starts_with and end_with with find.

I did this because LLVM renamed startswith to starts_with, so the only way to keep it compatible with LLVM 13 is to use .find instead 😅 I can add a // NOTE comment for this!

@mrexodia thanks for your contribution. This looks find with me overall, although I was wondering why you replaced starts_with and end_with with find.

I did this because LLVM renamed startswith to starts_with, so the only way to keep it compatible with LLVM 13 is to use .find instead 😅 I can add a // NOTE comment for this!

I understand. But I would prefer to keep the same semantics and have extra #ifdefs

I have managed to fix the CI in #1750, so once that is in (hopefully tomorrow), we will be able to rebase and hopefully merge this one.

Sure I can change it, but find("prefix") == 0 and starts_with do have the same semantics…

starts_with is faster because it doesn't keep looking for a match in the rest of the string.

@mrexodia you are right that the semantics for starts_with is preserved by the change, although I think this is not the case for ends_with. The latter can also be resolved, but I think starts/ends_with is more readable, so I'd prefer to keep that. (There is also the performance issue mentioned by @251, although perhaps less of an issue in this particular context). Thanks again!

Alright, I rebased on master and did the #ifdef for starts_with. I used LLVM 16, because that is where starts_with was introduced (see https://github.com/mrexodia/llvm-headers for a quick reference).

Thank you for your changes, @mrexodia. This looks good now, so let's merge it. Although we might need to get back to it once we add CI targets for LLVM 17, 18 and 19 (we are currently working on this).