r? @aturon

(rust_highfive has picked a reviewer for you, use r? to override)

Can you measure the metadata size before/after this patch? I'm a little worried about these being so rarely used but having a relatively significant cost in terms of metadata size.

Size of libcore.rlib:
before: 20,958,680 bytes
after: 26,644,788 byte
after with all ABIs except for "Rust" and "C" removed: 21,947,498 bytes
after with all ABIs except for "Rust" and "C" and functions with >5 arguments removed: 21,461,438 bytes

How about we stick to just Rust, C, and <5 arguments for now? I think we can also cut down on the number of traits implemented to the bare bones as well, I don't think these necessarily need to be at the same parity as raw pointers (e.g. are we really using function pointers as keys in hash maps?).

How about we stick to just Rust, C, and <5 arguments for now?

Good, it was the back-up plan from the beginning.

are we really using function pointers as keys in hash maps

We aren't, because they don't implement Hash!

I think we can also cut down on the number of traits implemented to the bare bones as well, I don't think these necessarily need to be at the same parity as raw pointer

I'd prefer not to do this, libcore size is not a holy grail after all. What increase in size of libcore do you consider acceptable?
Here's my chart of usefulness for these impls.
Clone (+ 150kb) - was already implemented
PartialEq/Eq (+ 300kb) - was partially implemented
Debug (+ 150kb) - needed if you want to assert_eq that PartialEq above or derive Debug for a containing struct.
PartialOrd/Ord (+ 300kb) or Hash (+ 150kb) are needed to have any kind of set of fn pointers or structs containing them (hash set, tree set, sorted array).
fmt::Pointer (+ 150kb) - needed mostly for consistency with data pointers.

OK, seems reasonable, I guess we can try to figure out how to control metadata size later.

@bors: r+ bc7c430

Thanks!

⌛ Testing commit bc7c430 with merge bf60883...

💔 Test failed - auto-win-gnu-64-opt

@bors: retry

On Mon, Sep 14, 2015 at 6:38 PM, bors notifications@github.com wrote:

[image: 💔] Test failed - auto-win-gnu-64-opt
http://buildbot.rust-lang.org/builders/auto-win-gnu-64-opt/builds/1409

—
Reply to this email directly or view it on GitHub
#28268 (comment).

⌛ Testing commit bc7c430 with merge 8320345...

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-x-android-t, auto-mac-32-opt, auto-mac-64-nopt-t, auto-mac-64-opt, auto-win-gnu-32-nopt-t, auto-win-gnu-32-opt, auto-win-gnu-64-nopt-t, auto-win-gnu-64-opt, auto-win-msvc-32-opt, auto-win-msvc-64-opt