make with_hasher_in const (consistent with with_hasher) #355
Conversation
|
rust-lang/rust#93706, seems that it depends a feature which was stabilized? |
|
You need to bump the minimum Rust version in CI to 1.61 which is when the |
I prefer to do that after 1.64 stabilized, just hold the PR. |
Signed-off-by: TennyZhuang <zty0826@gmail.com>
Signed-off-by: TennyZhuang <zty0826@gmail.com>
TennyZhuang
force-pushed
the
make-with-hasher-in-const
branch
from
d8a2ed1
to
bdf3e2d
Compare
|
Why were most actions canceled without errors? Are there some limitations? |
|
The runs are canceled once the first runner fails. In this case it seems that the 1.56.1 runner is failing, because you missed that one when changing the workflows (line 69). |
Signed-off-by: TennyZhuang <zty0826@gmail.com>
TennyZhuang
force-pushed
the
make-with-hasher-in-const
branch
from
bdf3e2d
to
630f3a7
Compare
|
Signed-off-by: TennyZhuang <zty0826@gmail.com>
|
Thanks! |
|
bors r+ |
|
@bors r+ |
make with_hasher_in const (consistent with with_hasher) I found that the following four methods have additional comments about their allocation: * HashMap::with_hasher * HashMap::with_hasher_in * HashSet::with_hasher * HashSet::with_hasher_in There is no reason to make such a difference; in this PR, I make them consistent as the following behavior: * All methods clarify that they will not allocate memory before the first insertion * All methods are marked as const function. We also bump the MSRV to 1.61.0 for the const_fn_trait_bound feature. Currently, 1.64.0 has been released, and I guess it's acceptable to break the 1.61.0 users. Signed-off-by: TennyZhuang <zty0826@gmail.com>
|
💔 Test failed - checks-actions |
|
I can't reproduce this failure locally, let's try again? |
make with_hasher_in const (consistent with with_hasher) I found that the following four methods have additional comments about their allocation: * HashMap::with_hasher * HashMap::with_hasher_in * HashSet::with_hasher * HashSet::with_hasher_in There is no reason to make such a difference; in this PR, I make them consistent as the following behavior: * All methods clarify that they will not allocate memory before the first insertion * All methods are marked as const function. We also bump the MSRV to 1.61.0 for the const_fn_trait_bound feature. Currently, 1.64.0 has been released, and I guess it's acceptable to break the 1.61.0 users. Signed-off-by: TennyZhuang <zty0826@gmail.com>
|
Actually this seems to be caused by: rust-lang/rust#102748 Let's wait for that fix to get merged first. |
|
💔 Test failed - checks-actions |
|
I think it can be merged now, but please don't bump the version of hashbrown crate because I want to document a raw part of crate |
|
@bors retry |
|
☀️ Test successful - checks-actions |
I found that the following four methods have additional comments about their allocation:
There is no reason to make such a difference; in this PR, I make them consistent as the following behavior:
We also bump the MSRV to 1.61.0 for the const_fn_trait_bound feature. Currently, 1.64.0 has been released, and I guess it's acceptable to break the 1.61.0 users.
Signed-off-by: TennyZhuang zty0826@gmail.com