I'm unsure what upstream LLVM does here and can't find any documentation so thought I would just ask for clarification.
In xdsl, some tests are failing with the latest release (see xdslproject/xdsl#6411). Some of these are #48, which I'm happy to change to CHECK-EMPTY, but some seem to be to do with --strict-whitespace.
It seems that the rule for checking the amount of whitespace is strange here, because:
- It seems 0 whitespace is counted as 0. i.e.
// CHECK:builtin.module { validates builtin.module {
- 1 whitespace is actually 0. i.e.
// CHECK: builtin.module { validates builtin.module {
- more generally
n+1 is actually n in the same way, so that the first space of the pattern seems to be ignored.
I can change the tests, but wanted to make sure that this was the intentional behaviour first. TBH it would make more sense to me to say that there always must be a space after the colon, because then at least its consistent.
I'm unsure what upstream LLVM does here and can't find any documentation so thought I would just ask for clarification.
In xdsl, some tests are failing with the latest release (see xdslproject/xdsl#6411). Some of these are #48, which I'm happy to change to
CHECK-EMPTY, but some seem to be to do with--strict-whitespace.It seems that the rule for checking the amount of whitespace is strange here, because:
// CHECK:builtin.module {validatesbuiltin.module {// CHECK: builtin.module {validatesbuiltin.module {n+1is actuallynin the same way, so that the first space of the pattern seems to be ignored.I can change the tests, but wanted to make sure that this was the intentional behaviour first. TBH it would make more sense to me to say that there always must be a space after the colon, because then at least its consistent.