librustc_*: Use `pub(crate)` instead of `pub` and remove dead code #43192

petrochenkov · 2017-07-12T18:47:59Z

This PR does the next things:

Replaces all pubs in librustc* crates (except for librustc itself) with pub(crate)s.
Reverts all replacements preventing x.py test from passing and RLS/Miri/Clippy from building.
Removes all dead code reported by the dead code lint.

The changes are 99% mechanical. I didn't judge and (semi-automatically) removed everything the compiler told me. Now this needs to be reviewed and some false positives need to be reverted (either for interface completeness or because they are needed by some third party tools).
EDIT: Changes mentioned in review comments are reverted.

cc @nikomatsakis and @michaelwoerister for rustc_data_structures
cc @nrc for librustc_driver APIs and tools
cc @rust-lang/compiler for everything else, if you see something that should be kept please leave a comment!

Next steps after reviewing:

Replace pub(crate) with nothing in crate roots.
Fix tidy and formatting.

rust-highfive · 2017-07-12T18:48:05Z

r? @eddyb

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

kennytm · 2017-07-12T19:05:40Z

src/librustc/lint/mod.rs

Just use :vis to combine all 3 arms here?

Ugh, it requires #![feature(macro_vis_matcher)] in all crates where the macro containing it is used, including tests.

eddyb · 2017-07-12T19:27:09Z

src/librustc_back/dynamic_lib.rs

Any chance a bunch of these are used on only some host OSes?

Looks like no, but CI should catch if I missed anything.

mcarton · 2017-07-12T20:23:20Z

We'd be lucky if something like that does not break Clippy. Cc @Manishearth, @llogiq, @oli-obk

oli-obk · 2017-07-13T07:28:18Z

This is great!

Would you be so kind and also test whether clippy and miri keep building and adjust your PR accordingly?

michaelwoerister

I did a pass over rustc_data_structures.

michaelwoerister · 2017-07-13T10:26:43Z

src/librustc_data_structures/bitvec.rs

Please keep this one.

michaelwoerister · 2017-07-13T10:28:25Z

src/librustc_data_structures/array_vec.rs

I think it makes sense to keep this type publicly available.

michaelwoerister · 2017-07-13T10:29:00Z

src/librustc_data_structures/accumulate_vec.rs

Please keep these pub.

michaelwoerister · 2017-07-13T10:29:42Z

src/librustc_data_structures/base_n.rs

Please keep this pub.

michaelwoerister · 2017-07-13T10:30:43Z

src/librustc_data_structures/bitvec.rs

Please keep pub.

michaelwoerister · 2017-07-13T10:41:38Z

src/librustc_data_structures/small_vec.rs

Please keep pub.

michaelwoerister · 2017-07-13T10:41:45Z

src/librustc_data_structures/snapshot_vec.rs

Please keep pub.

michaelwoerister · 2017-07-13T10:41:49Z

src/librustc_data_structures/snapshot_vec.rs

Please keep pub.

michaelwoerister · 2017-07-13T10:41:54Z

src/librustc_data_structures/snapshot_vec.rs

Please keep pub.

michaelwoerister · 2017-07-13T10:42:34Z

src/librustc_data_structures/tuple_slice.rs

This still seems like a useful type to have.

oli-obk · 2017-07-14T19:31:38Z

Thanks!

petrochenkov · 2017-07-14T20:22:12Z

Updated, comments addresses.

petrochenkov · 2017-07-14T21:01:07Z

src/librustc_allocator/lib.rs

is_unsafe being unused looks like a bug. It should be used in expansion, but it's ignored and all methods are expanded as unsafe.
@alexcrichton, is this intentional or just an omission?

Ah yes historically used but no longer used! Should be safe to remove.

nikomatsakis · 2017-07-17T15:03:41Z

I think these changes look good. I am sad to see some of the code go (e.g., useful algorithms on graphs etc), but I think that it's best to kill dead code. Those bits are still around in the git history if we ever want them. (In any case, I suspect we should start moving the librustc_data_structures code into individual crates; I have a local branch for example that removes the unification code in favor of https://crates.io/crates/ena.)

alexcrichton · 2017-07-20T18:30:40Z

ping r? @eddyb

eddyb · 2017-07-20T19:32:18Z

src/librustc_errors/lib.rs

Good riddance! 🎉

eddyb · 2017-07-20T19:40:46Z

src/librustc_platform_intrinsics/x86.rs

Was the generator for these updated?

Updated in the latest commit.

eddyb · 2017-07-20T19:41:17Z

src/librustc_privacy/lib.rs

Is this correct? Seems a mistake.

Ha, I fixed several similar cases, but haven't seen this one.
I grepped a bit harder and found a few more, fixed in the last commit.

eddyb · 2017-07-20T20:04:16Z

src/librustc_typeck/diagnostics.rs

Is this file some sort of rebase failure?

eddyb

Went through the entire diff and left some comments. Not that worried about deletions, because of git history, but some of the changes seemed unintentional.

nrc · 2017-07-20T21:53:43Z

src/librustc_save_analysis/lib.rs

These Data structures should all be pub still, they are expected to be used by clients

nrc · 2017-07-20T21:54:42Z

src/librustc_save_analysis/lib.rs

LIkewsie this trait and the functions below are API and should still be public

nrc · 2017-07-20T21:56:01Z

src/librustc_driver/driver.rs

These fields should still be pub

nrc · 2017-07-20T21:56:09Z

src/librustc_driver/driver.rs

I think this should be pub still

nrc · 2017-07-20T21:56:39Z

src/librustc_driver/driver.rs

All these changes from here on down are bogus and should still be pub.

nrc · 2017-07-20T21:58:10Z

src/librustc_driver/lib.rs

nrc · 2017-07-20T21:58:34Z

src/librustc_driver/lib.rs

nrc · 2017-07-20T21:59:11Z

src/librustc_driver/lib.rs

nrc · 2017-07-20T21:59:18Z

src/librustc_driver/lib.rs

nrc · 2017-07-20T22:00:07Z

src/librustc_errors/diagnostic.rs

pretty much everything here should be pub

where 'here' = all of the _errors crate

I made all the interface of Diagnostic public, but kept stuff in other modules containing implementation details crate-private.

nrc · 2017-07-20T22:07:03Z

I think that any change from pub -> pub(crate) in a non-public module is just noise (both in the git history and the source code). I think I'm generally a bit sad about how much noisier this is making the code, even if the tighter encapsulation and dead code removal have benefits.

bors · 2017-07-21T04:51:00Z

☔ The latest upstream changes (presumably #43183) made this pull request unmergeable. Please resolve the merge conflicts.

Revert some `pub(crate)`s to `pub`s to make sure Miri and Clippy work Revert some `pub(crate)`s to `pub`s to address Michael Woerister's comments

petrochenkov · 2017-07-21T22:20:57Z

Rebased, comments addressed.

petrochenkov · 2017-07-21T22:26:54Z

@nrc

I think that any change from pub -> pub(crate) in a non-public module is just noise

Interesting, this was kinda the point for me.
I don't immediately know if the module with pub entity private and even if it is private, the entity still can be reexported from it publicly. With pub(crate) the visibility scope becomes immediately evident.

bors · 2017-07-23T04:27:33Z

☔ The latest upstream changes (presumably #43387) made this pull request unmergeable. Please resolve the merge conflicts.

nrc · 2017-07-24T02:09:23Z

I don't immediately know if the module with pub entity private and even if it is private, the entity still can be reexported from it publicly. With pub(crate) the visibility scope becomes immediately evident.

This is pretty much a weakness in Rust's privacy system (and one the lang team has been debating recently). Does the privacy annotation on an item mean that item is 'at most' this public or 'at least' this public?

Using pub(crate) only works for doing what you want if you only care about privacy between crates. Many people care as much or more about privacy between modules and there using pub(crate) as an 'export limiter' is not helpful.

Given that this is only helpful for crate-level encapsulation and it is so noisey, I'd prefer not to do it.

arielb1 · 2017-07-25T23:23:32Z

Given that this is only helpful for crate-level encapsulation and it is so noisey, I'd prefer not to do it.

At least with pub(crate) you only have a limited amount of code that can use the item. Privacy levels between private and pub(crate) are only interesting when you have a complex in-crate module hierarchy, and I'll expect that in these cases inter-crate privacy will be somewhat dynamic anyway.

petrochenkov · 2017-07-25T23:41:20Z

this is only helpful for crate-level encapsulation

The primary benefit from pub(crate) compared to pub (in addition to encapsulation) is also lints catching various kinds of dead code - all of their logic is limited to a single crate and pub items are assumed to be used somewhere in other crates.

petrochenkov · 2017-07-25T23:44:13Z

This probably needs some collective decision from people working on these crates.
I see three primary alternatives:

The PR should be landed as is (after rebase)
Only the dead code removal part should be landed
The PR should be closed

nrc · 2017-07-26T07:06:49Z

I think the dead code removal is certainly worth landing. I would prefer not to land the rest, however, I don't work much on the crates which are mostly affected so if others want the changes, I won't object.

petrochenkov · 2017-07-26T08:57:22Z

change from pub -> pub(crate) ... is just noise ... in the git history

This annoys me too, btw.
I can squash all the commits to minimize the effect.
Additionally the formatting commit (b9b1013) can be reverted if history is considered more important than indent of function arguments.

aidanhs · 2017-08-03T18:37:56Z

@petrochenkov looks like this needs a rebase!

This probably needs some collective decision from people working on these crates.

Who are you referring to here? i.e. is this with a rust team to come to a decision?

petrochenkov · 2017-08-03T21:02:05Z

@aidanhs
The compiler team most likely.
A good occasion to try rfcbot (will it listen to me?).
EDIT: It won't listen :( Could someone with rights do this instead?

@rfcbot fcp merge

Questions:

Should the dead code removal part of this PR be merged?
Should the "pub -> pub(crate)" part of this PR be merged?
Should the formatting commit be reverted to reduce the noise in git history?

eddyb · 2017-08-03T21:06:50Z

@rfcbot fcp merge

rfcbot · 2017-08-03T21:59:39Z

Team member @eddyb has proposed to merge this. The next step is review by the rest of the tagged teams:

No concerns currently listed.

Once these reviewers reach consensus, this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

See this document for info about what commands tagged team members can give me.

petrochenkov · 2017-08-19T11:28:28Z

Too much bit rot and not enough interest, closing.

@eddyb

rustc: Remove some dead code Extracted from #43192 r? @eddyb

rust-highfive assigned eddyb Jul 12, 2017

kennytm reviewed Jul 12, 2017

View reviewed changes

petrochenkov changed the title ~~librustc_*: Use pub(crate) instead of pub and remove dead code~~ [WIP] librustc_*: Use pub(crate) instead of pub and remove dead code Jul 12, 2017

eddyb reviewed Jul 12, 2017

View reviewed changes

aidanhs added the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Jul 13, 2017

michaelwoerister reviewed Jul 13, 2017

View reviewed changes

petrochenkov force-pushed the pubcrate branch 2 times, most recently from 265ae78 to 55694ca Compare July 14, 2017 20:21

petrochenkov commented Jul 14, 2017

View reviewed changes

petrochenkov force-pushed the pubcrate branch 2 times, most recently from 9b1c051 to 9153114 Compare July 14, 2017 23:39

alexcrichton added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Jul 20, 2017

eddyb reviewed Jul 20, 2017

View reviewed changes

src/librustc_errors/lib.rs Outdated

Copy link

Member

eddyb Jul 20, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good riddance! 🎉

eddyb reviewed Jul 20, 2017

View reviewed changes

src/librustc_typeck/diagnostics.rs Outdated

Copy link

Member

eddyb Jul 20, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this file some sort of rebase failure?

eddyb suggested changes Jul 20, 2017

View reviewed changes

nrc suggested changes Jul 20, 2017

View reviewed changes

petrochenkov added 2 commits July 21, 2017 23:29

Use pub(crate) instead of pub where possible in librustc_* crates

532bb60

Remove dead code

dc6b8ed

Revert some `pub(crate)`s to `pub`s to make sure Miri and Clippy work Revert some `pub(crate)`s to `pub`s to address Michael Woerister's comments

Address more comments

8898cdb

petrochenkov force-pushed the pubcrate branch from 9153114 to 8898cdb Compare July 21, 2017 22:20

eddyb approved these changes Jul 21, 2017

View reviewed changes

petrochenkov changed the title ~~[WIP] librustc_*: Use pub(crate) instead of pub and remove dead code~~ librustc_*: Use pub(crate) instead of pub and remove dead code Jul 23, 2017

arielb1 added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jul 25, 2017

rfcbot added the proposed-final-comment-period Proposed to merge/close by relevant subteam, see T-<team> label. Will enter FCP once signed off. label Aug 3, 2017

petrochenkov mentioned this pull request Aug 19, 2017

rustc: Remove some dead code #43986

Merged

petrochenkov closed this Aug 19, 2017

rfcbot removed the proposed-final-comment-period Proposed to merge/close by relevant subteam, see T-<team> label. Will enter FCP once signed off. label Aug 19, 2017

bors added a commit that referenced this pull request Aug 21, 2017

Auto merge of #43986 - petrochenkov:pubcrate3, r=pnkfelix

757b7ac

rustc: Remove some dead code Extracted from #43192 r? @eddyb

petrochenkov mentioned this pull request Sep 1, 2017

RFC: Clarify and streamline paths and visibility rust-lang/rfcs#2126

Merged

petrochenkov deleted the pubcrate branch June 5, 2019 15:51

librustc_*: Use pub(crate) instead of pub and remove dead code #43192

librustc_*: Use pub(crate) instead of pub and remove dead code #43192

Uh oh!

Conversation

petrochenkov commented Jul 12, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rust-highfive commented Jul 12, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mcarton commented Jul 12, 2017

Uh oh!

oli-obk commented Jul 13, 2017

Uh oh!

michaelwoerister left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

oli-obk commented Jul 14, 2017

Uh oh!

petrochenkov commented Jul 14, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

nikomatsakis commented Jul 17, 2017

Uh oh!

alexcrichton commented Jul 20, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

eddyb left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

librustc_*: Use `pub(crate)` instead of `pub` and remove dead code #43192

librustc_*: Use `pub(crate)` instead of `pub` and remove dead code #43192

petrochenkov commented Jul 12, 2017 •

edited

Loading

arielb1 commented Jul 25, 2017 •

edited

Loading