reduce dependencies for windows-msvc targets

[klensy]

Some dependencies don't used on windows-msvc targets, so exclude them:

• miniz_oxide (+ adler)
• addr2line (+ gimli)
• object (+ memchr)
• libc too

Will sync with rust repo

[klensy]

Should anything be fixed in https://github.com/rust-lang/backtrace-rs/blob/master/crates/as-if-std/Cargo.toml?

Looking at

backtrace-rs/src/symbolize/mod.rs

Lines 466 to 485 in e1c49fb

	cfg_if::cfg_if! {
	if #[cfg(miri)] {
	mod miri;
	use miri as imp;
	} else if #[cfg(all(windows, target_env = "msvc", not(target_vendor = "uwp")))] {
	mod dbghelp;
	use dbghelp as imp;
	} else if #[cfg(all(
	any(unix, windows),
	not(target_vendor = "uwp"),
	not(target_os = "emscripten"),
	any(not(backtrace_in_libstd), feature = "backtrace"),
	))] {
	mod gimli;
	use gimli as imp;
	} else {
	mod noop;
	use noop as imp;
	}
	}

gimli module never used on windows-msvc targets?

addr2line used only in gimli module which is unused on windows-msvc and backtrace_in_libstd feature?

[workingjubilee]

I believe this would/should address this issue:

• Make gimli / addr2line optional on windows #451

So, yes.

[klensy]

Should anything be fixed in https://github.com/rust-lang/backtrace-rs/blob/master/crates/as-if-std/Cargo.toml?

Last question left.

[workingjubilee]

So I believe they should have matching cfgs for their deps, especially if that would allow us to build less deps while building std? But perhaps I am wrong somehow?

[philipc]

Yes, the deps in as-if-std/Cargo.toml must match the main Cargo.toml. They are only duplicated there because there's no other way to do it, not because they might be different.

[klensy]

Okay, looks like it ready.

[klensy]

nano ping.

[klensy]

Can this be reviewed, so able to be merged in rust repo? Thanks.

[workingjubilee]

This is still on my radar, my most recent agenda has been pushing a few PRs that were opened in 2022(!) or even 2021(!) over the edge now that some of them have been unblocked.

[workingjubilee]

Blech, merge conflicts...

[workingjubilee]

Sorry for not getting back to this earlier!

[workingjubilee]

I would prefer we keep this as an "exploded-out" table here, as recent events have demonstrated that: it is extremely likely this will be diffed and that it may only diff a single field, so I would prefer to minimize the number of merge conflicts we have. I realize this means it looks like

[target.'cfg(not(all(windows, target_env = "msvc", not(target_vendor = "uwp"))))'.dependencies.object]
version = "0.32.0"
default-features = false
optional = true
features = ['read_core', 'elf', 'macho', 'pe', 'unaligned', 'archive']

[klensy]

For some reason, optional = true exist there, but not in top Cargo.toml

[workingjubilee]

Same remark, this should be

[target.'cfg(not(all(windows, target_env = "msvc", not(target_vendor = "uwp"))))'.dependencies.object]
version = "0.32.0"
default-features = false
optional = true
features = ['read_core', 'elf', 'macho', 'pe', 'unaligned', 'archive']

[klensy]

No optional = true here.

[workingjubilee]

Oh true.

[workingjubilee]

Nice!

[klensy]

For some reason CI have troubles downloading syn for windows.

[ChrisDenton]

syn has been causing a lot of intermittent issues in CI lately, across a number of different repositories.

Edit: well I don't know if syn is the root cause or just a common symptom.

[workingjubilee]

It is possible that the author's recent choice to ship precompiled binaries for certain relevant proc-macros is taxing various points of network IO more severely: serde-rs/serde#2538

[pacak]

is taxing various points of network IO more severely

The binary in question is 680kb compresses down to 300kb. Shipping a binary is not the best idea, but I don't think it should affect network that much.

[workingjubilee]

Yah, dtolnay thinks it's a cargo bug and I think he's right.

[klensy]

Funny thing, now i see this here: https://github.com/rust-lang/rust/actions/runs/5887880758/job/15968022846?pr=114923

Thu, 17 Aug 2023 07:15:39 GMT       Updating crates.io index
Thu, 17 Aug 2023 07:16:10 GMT   warning: spurious network error (3 tries remaining): [28] Timeout was reached (Operation too slow. Less than 10 bytes/sec transferred the last 30 seconds)
Thu, 17 Aug 2023 07:16:40 GMT   warning: spurious network error (2 tries remaining): [28] Timeout was reached (Operation too slow. Less than 10 bytes/sec transferred the last 30 seconds)
Thu, 17 Aug 2023 07:17:14 GMT   warning: spurious network error (1 tries remaining): [28] Timeout was reached (Operation too slow. Less than 10 bytes/sec transferred the last 30 seconds)
Thu, 17 Aug 2023 07:17:50 GMT   error: failed to get `zerofrom-derive` as a dependency of package `zerofrom v0.1.2`
Thu, 17 Aug 2023 07:17:50 GMT       ... which satisfies dependency `zerofrom = "^0.1.1"` (locked to 0.1.2) of package `icu_provider v1.2.0`
Thu, 17 Aug 2023 07:17:50 GMT       ... which satisfies dependency `icu_provider = "^1.2.0"` (locked to 1.2.0) of package `icu_list v1.2.0`
Thu, 17 Aug 2023 07:17:50 GMT       ... which satisfies dependency `icu_list = "^1.2"` (locked to 1.2.0) of package `rustc_baked_icu_data v0.0.0 (/checkout/compiler/rustc_baked_icu_data)`
Thu, 17 Aug 2023 07:17:50 GMT       ... which satisfies path dependency `rustc_baked_icu_data` (locked to 0.0.0) of package `rustc_error_messages v0.0.0 (/checkout/compiler/rustc_error_messages)`
Thu, 17 Aug 2023 07:17:50 GMT       ... which satisfies path dependency `rustc_error_messages` (locked to 0.0.0) of package `rustc_driver_impl v0.0.0 (/checkout/compiler/rustc_driver_impl)`
Thu, 17 Aug 2023 07:17:50 GMT       ... which satisfies path dependency `rustc_driver_impl` (locked to 0.0.0) of package `rustc-main v0.0.0 (/checkout/compiler/rustc)`