In D71524#2291925, @steakhal wrote:

In D71524#2284386, @Szelethus wrote:

I figured you're still working on this, sorry! I'd really like to chat about my earlier comment D71524#1917251, as it kind of challenges the high level idea.

What about marking the std::cin object itself as tainted and any object created by ifstream::ifstream(const char*) or similar functions.
Then propagate taint via the extraction operator (operator>>) only if the stream was tainted.
This way we could reduce the false-positives of this crude heuristic. What do you think?

Ping

Rebase to master.

@steakhal's revision is on the top of this. Changing the order will only cause unnecessary work on both sides.

Ping

I did the required changes and tried to commit it, but I couldn't. I heard the codebase was migrated to GitHub. Maybe it affected my commit access.

Ping

Ping

Resolving the problems properly seemed quite difficult. I just finished university and started to work, therefore I don't have enough free time to finish the checker. Sorry about that.

Closed by 080ecafdd8b3e990e5ad19202d089c91c9c9b164.

Should I do anything or it is ready?

I think it shouldn't give compile error in case of incorrect configuration now (maybe warning) because:

Report diagnostic error in case of an invalid filename or an ill-formed yaml file.

I already thought about it. It would make the code much cleaner, but it would have a little performance impact (Does it matter?).
It's straightforward to read the supported functions from another yaml file. Besides that, it can support multiple config files too.

I changed the parsing, therefore the return value index is represented by -1.
I added a test configuration file and parse it when testing to prove the configuration doesn't break the code.

Sorry for the late answer, I was working on my thesis which is about taint analysis. During that, I implemented several cool features (namespaces, std::cin, std::string, etc.) for the checker. I will share them soon.

Rebase after https://reviews.llvm.org/D59861.
Fix custom filter test case: functions without definition always remove taintedness.

Rebase after https://reviews.llvm.org/D59861.

Rebase after https://reviews.llvm.org/D59861.

Thanks, it will make my changes more cleaner.

I add a new taint type, which represents a lack of taintedness. That's why I changed the name of addTaint() to setTaint(). Of course, it's not an important change, I can move it to another patch.

Why is it better not to use static functions/variables? Has it any performance impact?

Yes, they do. Thanks for the review.

Yes, thank you.

Is it ready to land?

I don't have commit access. Can you please do it on my behalf?

Rebase patch on the earlier version. Remove unnecessary inline specifiers. Fix typo.

I use std::move which is in the <utility>.

Upload diff with full context.

If you find it ok, can you please commit it? @xazax.hun

How can I commit it?

Add std::srand check to C++ tests. Format patch.

I think I resolved all of the comments. Do I forget anything?

Add capability to provide a user-configurable list of disallowed types which will trigger warning. Now it also detects C srand function. Rename check to cert-msc51-cpp. Add cert-msc32-c check. Change warning messages. Other minor fixes.