Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CI considers [Info] messages to be a violation of //NOWARN #1151

Open
michael-schwarz opened this issue Sep 6, 2023 · 4 comments
Open

CI considers [Info] messages to be a violation of //NOWARN #1151

michael-schwarz opened this issue Sep 6, 2023 · 4 comments
Labels
bug setup Dependencies, CI, releasing testing

Comments

@michael-schwarz
Copy link
Member

In the course of #1094, we discovered that our CI considers messages such as the following to be a violation of the //NOWARN annotation.

[Info][Imprecise] Invalidating expressions: AddrOf(Var(tmp___0, NoOffset)) (05-oob-implicit-deref.c:10:4-10:35)

Normally, the solution might be to not write such annotations, but the problem is that such issues may, e.g., arise from a high fortification level, turning

// This     
memset(ptr, 0, four * sizeof(int)); //NOWARN 

// into this
size_t  tmp = __builtin_object_size(ptr, 0); //NOWARN
__builtin___memset_chk(ptr, 0, four * sizeof(int), tmp); //NOWARN

where with //NOWARN we want to check the absence of some actual warning rather than some internal.

My proposed solution would be to not consider messages of the severity Info as violations of the //NOWARN annotation. Warnings should be given to the user with a higher severity anyway.

@sim642
Copy link
Member

sim642 commented Sep 8, 2023

At some point this was required because // WARN doesn't allow more precise matching.

This on its own won't really fix the issues with fortification though. There's #706 about it. The trouble is that with fortification some warnings move to header files where our comment annotations no longer check them correctly.

@michael-schwarz
Copy link
Member Author

Yes, it won't fix the fortification issue in general, but it would fix this specific issue.

@sim642
Copy link
Member

sim642 commented Sep 8, 2023

These are the test failures for // WARN when changing this:

Expected warn, but registered other on assume-unknown:16
  __goblint_assume_join(id2); // WARN joining unknown thread ID, make joined set All threads
Expected warn, but registered other on assume-unknown:20
  pthread_create(&id, NULL, t_fun, NULL); // WARN make joined set different from All threads
51/05 failed!

Expected warn, but registered other on mixedjmpbufs:12
    longjmp(error1, 1); // WARN (modified since setjmp)
Expected warn, but registered other on mixedjmpbufs:20
    longjmp(error0, 1); // WARN (modified since setjmp)
68/31 failed!

Expected warn, but registered other on munge:41
      longjmp(buf1, 1); //WARN
68/33 failed!

3 test(s) failed: ["51/05 assume-unknown", "68/31 mixedjmpbufs", "68/33 munge"]

For // NOWARN there's no good way to know where it might be tested for.

@michael-schwarz
Copy link
Member Author

The question is whether some other message is simply hiding a bug here... We should produce actual warnings in those cases and not just some random info messages....

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug setup Dependencies, CI, releasing testing
Projects
None yet
Development

No branches or pull requests

2 participants