On Saturday, October 19, 2019 at 1:20:15 PM UTC-7, David Roe wrote: > > I think the idea is that if a user's typing interactively then they don't > need a deprecation warning (since the behavior currently works). It's more > important to show a user a warning if they have the deprecated behavior in > a function they've written. > I'm pretty sure that's not the reason. As you can see from https://trac.sagemath.org/ticket/28500, the stack depth is tuned so that the warning references the stack frame of the *call site* of the function raises the deprecation warning. That makes sense, because that's where the change should be made if you want to avoid calling a deprecated function. That works just fine from the interpreter top level, because there is a stack frame corresponding to top level.
The problem arises when the deprecation warning is raised in a cython function, since cython doesn't create python stack frames. With one less frame on the stack, the warning ends up referencing the level *above* the call site, which is there in the doctest framework, but not at top-level in the interpreter. In the doctest it means there's a nonsensical callsite referenced (but this is not checked by the doctests); in the interpreter it means the warning gets swallowed because a a failure in finding a stack frame [I think that behaviour is a bug in Python's warning machinery. It's a silent failure to issue a warning in a case where some warning is almost certainly desired]. The obvious solution is to use a different "deprecation" in cython code: one that uses a stack depth of one less. This is quite easy to implement by providing a "deprecation_cython" as well and use something like "from sage.misc.superseded import deprecation_cython as deprecation" in cython files instead of the usual. Making this change is quite straightforward (it can probably be done by a script if you really want to) and is completely uncontroversial. It is a bit of work, though, and it will be touching a lot of files. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/9f02b869-2528-436e-96c3-dcabe792d32b%40googlegroups.com.
