Update cargo

[ehuss]

13 commits in e515c3277bf0681bfc79a9e763861bfe26bb05db..33ee5f82edb50af87b952c5b28de0f5fb41ebf18
2021-09-08 14:32:15 +0000 to 2021-09-17 13:51:54 +0000

• Update curl-sys (Update curl-sys cargo#9917)
• Bump Cargo's curl requirement to 7.79.0 (Bump Cargo's curl requirement to 7.79.0 cargo#9914)
• Revert "When a dependency does not have a version, git or path, fails directly" (Revert "When a dependency does not have a version, git or path, fails directly" cargo#9911)
• Add some contributor docs for debugging testsuite tests. (Add some contributor docs for debugging testsuite tests. cargo#9904)
• Fix warnings when documenting with --document-private-items (Fix warnings when documenting with --document-private-items cargo#9903)
• Improve "wrong output" error. (Improve "wrong output" error. cargo#9905)
• Fix warnings from better precision of dead_code lint (Fix warnings from better precision of dead_code lint cargo#9906)
• Bump to 0.58.0, update changelog (Bump to 0.58.0, update changelog cargo#9900)
• Fix rustc --profile=dev unstable check. (Fix rustc --profile=dev unstable check. cargo#9898)
• config.md: fix typo (config.md: fix typo cargo#9896)
• Enable some tests on windows. (Enable some tests on windows. cargo#9893)
• Enable strip test on macos. (Enable strip test on macos. cargo#9889)
• Fix cargo fix --edition on stable. (Fix cargo fix --edition on stable. cargo#9890)

[rust-highfive]

⚠️ Warning ⚠️

• These commits modify submodules.

[ehuss]

@bors r+

[bors]

📌 Commit 4be29a871ea86dcf973eaee0227837163009a6ad has been approved by ehuss

[ehuss]

@bors r-

[ehuss]

@bors r+ p=1

[bors]

📌 Commit d5307604d0e92663c527a7bed9b9fc190e28bb4e has been approved by ehuss

[bors]

🌲 The tree is currently closed for pull requests below priority 100. This pull request will be tested once the tree is reopened.

[bors]

⌛ Testing commit d5307604d0e92663c527a7bed9b9fc190e28bb4e with merge 4f31c8fd62cbae7b8912b7598f5232587348efa0...

[bors]

💔 Test failed - checks-actions

[alexcrichton]

Continuing this here since this may be the best place for this.

Looking into this I'm confused what's happening. The undefined symbols are in Curl_auth_create_digest_md5_message and it includes symbols from bufref.c and also vauth.c. The vauth.c is built (via build.rs) on Windows and for the spnego feature (neither of which I think is applicable here). The bufref.c is only built on Windows. This means that indeed the build script isn't building the symbols required by Curl_auth_create_digest_md5_message. That being said the md5 function is only called by one other function in curl, Curl_sasl_continue, which is part of curl_sasl.c which is never built by build.rs. This means that there shouldn't actually be any dependencies on Curl_auth_create_digest_md5_message which is probably why this works on other targets (the --gc-sections linker argument coupled with the -ffunction-sections argument should ensure that Curl_auth_create_digest_md5_message is gc'd away and we ignore its unresolved symbols).

This means that the bug here could be in the illumos-specific toolchain. I see that gcc 8.4.0 is being used with binutils 2.25.1. Those aren't really all that old and I just built and ran tests for curl in a ubuntu:14.04 container which uses gcc 4.8.4 and binutils 2.24 and everything worked out ok. My best guess at this point is that maybe one of the illumos-specific libraries references curl symbols and expects to get the system curl but ends up getting our more-incomplete bundled version. I... don't know why that would happen though.

I'll try to build the illumos container locally with this PR and dig further.

[alexcrichton]

Well silly me should have checked this first. The --gc-sections flag isn't passed on illumos due to its configuration of linker_is_gnu: false. I'll go try to add CI for this in the curl crate.

[ehuss]

@bors r+

[bors]

📌 Commit 1626fee has been approved by ehuss

[bors]

⌛ Testing commit 1626fee with merge 9dd4ce8...

[bors]

☀️ Test successful - checks-actions
Approved by: ehuss
Pushing 9dd4ce8 to master...

[Mark-Simulacrum]

For any perf triagers noticing the bootstrap and wall time improvement from this PR, it is almost certainly due to enabling performance rather than on demand cpu frequency governor on the perf collector. I think this got disabled by accident at some point (likely after a reboot).