-
Notifications
You must be signed in to change notification settings - Fork 12.7k
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
Fix bug in integer range matching #57978
Fix bug in integer range matching #57978
Conversation
(rust_highfive has picked a reviewer for you, use r? to override) |
r? @oli-obk |
}; | ||
if let Some((min, max, sz)) = range { | ||
if let (Some(lo), Some(hi)) = (lo.val.try_to_bits(sz), hi.val.try_to_bits(sz)) { | ||
let (lo, hi) = (lo ^ bias, hi ^ bias); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So this basically makes sure that unsigned comparison of signed numbers still works out?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah. I'll add a comment.
The job Click to expand the log.
I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
@bors r+ |
}; | ||
if let Some((min, max, sz)) = range { | ||
if let (Some(lo), Some(hi)) = (lo.val.try_to_bits(sz), hi.val.try_to_bits(sz)) { | ||
// We want to compare ranges numerically, but the order of signed floating- |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Signed integers
📌 Commit 441355d7c2fe99b1e666d3c30ee9b6c7996547bf has been approved by |
@bors r- r=me with the comment nit addressed |
@varkor your update to the comment (890198445cde6603ea58c67a3177d38c36da4ede) continues to refer to floating-point numbers rather than signed integers (which is what I had interpreted @matthewjasper 's comment nit as referring to). So now I am inferring that you did in fact mean to reference floating-point numbers in your comment. I haven't attempted to validate the reasoning in the comment/code when applied to floating-point numbers.
But, both your own code here and the referenced code in pattern/_match.rs, is only going to look at ranges constructed from one of the types { ... so ... is there some way that floating point values can arise in these code paths? Even the underlying representation in the MIR interpreter, as far as I can tell, avoids using floating-point values... |
(If in fact floating-point values can arise ... I guess I'd expect to see a unit test illustrating that scenario...) |
Sorry, I must have been distracted when editing the comment; it should not refer to floating-point numbers at all. I think I was conflating it with another problem I've been thinking about recently. Floating-point range patterns are not permitted in this location and do not need to be handled. I shall update the comment (and squash all of the edits) once I get back to my computer. Sorry for the confusion. |
8901984
to
cd1047e
Compare
@bors r=oli-obk |
📌 Commit cd1047e has been approved by |
…oli-obk Fix bug in integer range matching Fixes #57894.
☀️ Test successful - checks-travis, status-appveyor |
[beta] Rollup backports Cherry-picked: * #58008: Pass correct arguments to places_conflict * #58007: Don't panic when accessing enum variant ctor using `Self` in match * #57978: Fix bug in integer range matching * #57862: Build the standard library for thumbv7neon-unknown-linux-gnueabihf in CI * #57659: Fix release manifest generation r? @ghost
Fixes #57894.