You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
here is the second issue I found, where filecheck behaves differently from the official LLVM FileCheck.
In my project, I sometimes have multiple tests in the same file, like this:
RUN: echo "1" | filecheck %s --check-prefix FIRST
FIRST: 1
RUN: echo "2" | filecheck %s --check-prefix SECOND
SECOND: 2
This works completely fine. However, once one of the check-prefixes is a suffix of another check-prefix, filecheck want to check both, while LLVM FileCheck still works correctly.
Interesting! I had briefly considered that interaction previously, but didn't check how LLVM did it. I'll look into LLVM to see how they decide what a prefix actually is!
This PR adds the requirement that the `CHECK` be preceded by either a
new-line, or a character that is not `[A-Za-z]`. This fixes#28
It also adds diagnostic printing when checking fails inside a DAG region
by printing all currently considered input and greying out regions that
are already consumed.
Hi,
here is the second issue I found, where filecheck behaves differently from the official LLVM FileCheck.
In my project, I sometimes have multiple tests in the same file, like this:
This works completely fine. However, once one of the check-prefixes is a suffix of another check-prefix, filecheck want to check both, while LLVM FileCheck still works correctly.
I.e., if I change the test to
then filecheck fails with:
while LLVM FileCheck correctly sees SECONDTEST as a different prefix from TEST.
This can be relevant e.g. if you have tests like "INT_TEST" and "UINT_TEST" :)
The text was updated successfully, but these errors were encountered: