rust 1.52.1

[Bo98]

Created with brew bump-formula-pr.

[EricFromCanada]

Heads-up that the SDKROOT block can be removed. Its absence allows the current version to build on 10.12 and doesn't appear to affect later versions.

[carlocab]

1.52.1 has been released.

[Bo98]

Well you still have to fix the breakages caused by 1.52, if you're happy to do that go ahead. I won't be able to look into it until tomorrow.

[disrupted]

Well you still have to fix the breakages caused by 1.52, if you're happy to do that go ahead. I won't be able to look into it until tomorrow.

I see what you mean now. Was hoping for a quick fix to get the update for rust-analyzer through. But in that case I am going to close my PR as I think it will make more sense to make the necessary changes directly here if you can bump to v1.52.1

[carlocab]

Was hoping for a quick fix to get the update for rust-analyzer through.

Rust takes two days to run in CI, so the version bump was never going to be quick, unfortunately.

[cho-m]

Starting some initial failure analysis for Intel nodes.

Rust failures (i.e. may be due to formula bump):

• angle-grinder - actual Rust compilation failure. probably need to analyze further:

     Compiling ag v0.16.0 (/private/tmp/angle-grinder-20210506-90401-nbddyx/angle-grinder-0.16)
  error: proc-macro derive panicked
    --> src/alias.rs:13:26
     |
  13 | const ALIASES_DIR: Dir = include_dir!("aliases");
     |                          ^^^^^^^^^^^^^^^^^^^^^^^
     |
     = help: message: assertion failed: source.starts_with(prefix)
     = note: this error originates in a macro (in Nightly builds, run with -Z macro-backtrace for more info)

Failures probably not specific to this formula:

• asuka, vice, gifski, ktmpl, wasm-pack failures look identical to previous update rust 1.51.0 rust 1.51.0 #73924, so probably can ignore
• pdfpc - same upstream issue with Vala 0.52 (I opened Fix build with vala 0.52+ due to API change in gstreamer pdfpc/pdfpc#594)
• cargo-watch - reoccurring test failure over multiple Formula.
• caffe - test had another intermittent failure. Really need to find a way to fix this test case as it logs way too much.
• freeswitch freeswitch: fix build error for undeclared identifier 'NSIG' #77004 - related to upstream Make fails: libapr-1 error undeclared identifier 'NSIG' extern __const char *__const sys_signame[NSIG]; missing define in apr_general.h signalwire/freeswitch#1145.
• chronograf chronograf: fix build issue from incompatible [email protected] and node@16 #77018 - issue I mentioned for go 1.16.4 PR related to building with node 16
• ppsspp ppsspp: fix build with FFmpeg 4.4+ #77003 - issue due to newer ffmpeg. Probably will fail until a newer version is added to Homebrew with Remove deprecated API calls for new FFmpeg 4.3.x hrydgard/ppsspp#14176
• mlpack mlpack: update stb headers #76835 - fixed

[carlocab]

vice is a test-bot bug. asuka is only on Big Sur and doesn't have a Big Sur bottle, so we can ignore that one.

Does hrydgard/ppsspp#14176 apply cleanly to the current version in Homebrew/core?

[cho-m]

@carlocab probably worth a try in PR as I was able to git apply ... on top of git checkout v1.11.3

[carlocab]

This is the freeswitch error:

In file included from locks/unix/thread_mutex.c:17:
In file included from /private/tmp/freeswitch-20210507-44211-eeq084/libs/apr/include/arch/unix/apr_arch_thread_mutex.h:27:
In file included from /private/tmp/freeswitch-20210507-44211-eeq084/libs/apr/include/arch/unix/apr_private.h:881:
In file included from /private/tmp/freeswitch-20210507-44211-eeq084/libs/apr/include/arch/unix/../apr_private_common.h:24:
In file included from ./include/apr_pools.h:39:
In file included from ./include/apr_general.h:33:
/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/include/signal.h:69:42: error: use of undeclared identifier 'NSIG'
extern __const char *__const sys_signame[NSIG];
                                         ^
/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/include/signal.h:70:42: error: use of undeclared identifier 'NSIG'
extern __const char *__const sys_siglist[NSIG];
                                         ^

Logs: freeswitch.tar.gz

[cho-m]

Then the freeswitch failure may be the same issue discussed in closed attempt to bump to 1.10.6 (#75728) relating to upstream failure from signalwire/freeswitch#1145.

[carlocab]

Is freeswitch building its own apr? Maybe we can persuade it to use the brewed one, if it is?

[cho-m]

Not too sure, but configure.ac apr related section appears to be:
https://github.com/signalwire/freeswitch/blob/v1.10.5/configure.ac#L835-L838

AC_CHECK_LIB(apr-1, apr_pool_mutex_set, use_system_apr=yes, use_system_apr=no)
AM_CONDITIONAL([SYSTEM_APR],[test "${use_system_apr}" = "yes"])
AC_CHECK_LIB(aprutil-1, apr_queue_pop_timeout, use_system_aprutil=yes, use_system_aprutil=no)
AM_CONDITIONAL([SYSTEM_APRUTIL],[test "${use_system_aprutil}" = "yes"])

Probably need to check Homebrew's apr has apr_pool_mutex_set.

EDIT: Doesn't seem to be available in headers for Homebrew's or CLT's version of apr.
May be a function added into a forked apr
freeswitch code: https://github.com/signalwire/freeswitch/blob/v1.10.5/libs/apr/include/apr_pools.h#L427-L428

[cho-m]

For chronograf (and similar Node 16 failures from #76791 (comment)) is it better to:

• leave build failure?
• downgrade to node@14
• try to yarn upgrade [email protected] (only version that currently supports Node 16. The risk is that chronograf uses [email protected] so it is a pretty large upgrade).

[SMillerDev]

For chronograf (and similar Node 16 failures from #76791 (comment)) is it better to

I'd say downgrade node

[cho-m]

angle-grinder does --build-from-source with Rust 1.51.0.

So, that failure may actually be related to bump to Rust 1.52

[Bo98]

Yeah, I have a patch for that I'll test later today. I think it requires bumping proc-macro-hack from 0.4.2 to 0.4.3.

The rest aren't really blockers since they can be proven to not be Rust-related, but of course fixing them is ideal.

[carlocab]

Let's give this another go.

[cho-m]

angle-grinder is going to still fail to compile with Rust 1.52 since the lock file hasn't changed.

I guess there is an option to ignore failure for now and deal with it afterward.

[carlocab]

ARM failures with bottles:

• angle-grinder - error: cannot find macro proc_macro_call in this scope
• cargo-audit - curl: (22) The requested URL returned error: 404
• cargo-watch- common test failure (e.g. seen in [email protected] 3.9.5 #76545 rust: set MacOS env variables only on macOS #78107)
• gifski - issues with ffmpeg dependency. also, has been failing updates gifski 1.4.0 #69616, gifski 1.4.3 #73583
• gnumeric - gnumeric: revision bump (Big Sur Perl) #78444
• graph-tool - boost-related
• mpv - previously seen mpv: build for Apple Silicon #68460 (Build failure on Apple Silicon mpv-player/mpv#8445), but not easily reproducible
• pc6001vx - #include <version> bug

[carlocab]

Ok, the mpv build is just haunted

[carlocab]

Mojave

Error: 13 failed steps!
brew install --build-from-source angle-grinder
brew test --retry --verbose buku
brew install --build-from-source cargo-audit
brew test --retry --verbose cargo-watch
brew test --retry --verbose cargo-watch
brew install --build-from-source gifski
brew install --build-from-source graph-tool
brew test --retry --verbose ktmpl
brew install --build-from-source openimageio
brew install --build-from-source pdfpc
brew install --build-from-source vice
brew test --retry --verbose wasm-pack
brew test --retry --verbose wasm-pack

[cho-m]

Ignoring some standard failures and those analyzed for ARM, the remaining ones for Intel are:

• openimageio (All, build)
• buku (Mojave, test)
• deno (Big Sur, build) - LLVM ERROR: IO failure on output stream: No space left on device

[carlocab]

Can't reproduce the openimageio failure from current master.

Could it be a fluke? I don't think openimageio even needs rust at runtime.

[cho-m]

Not sure why openimageio started failing. It didn't fail in the previous run, but it did fail in the Linux-specific PR #78107, which didn't touch the Rust version.

It's possibly due to something else changing in build environment (e.g. other formulae or CI settings).

There is a Homebrew error on possible token expiration, though I haven't checked if this is cause of failure:

/usr/local/Homebrew/Library/Homebrew/utils/github/api.rb:290:in `raise_error': GitHub API Error: Bad credentials (GitHub::API::AuthenticationFailedError)
HOMEBREW_GITHUB_API_TOKEN may be invalid or expired; check:
  https://github.com/settings/tokens
	from /usr/local/Homebrew/Library/Homebrew/utils/github/api.rb:234:in `open_rest'
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:166:in `search'

[carlocab]

Yea, seems like a network error. Probs didn't get to do git clone properly or something.

We could try re-running to see if it was just a fluke. That's pretty costly, though. And I doubt any change in rust can actually break openimageio without also breaking the intermediate formulae that connect openimageio to rust in the dependency graph.

[cho-m]

I also didn't see an issue building both rust 1.52.1 and openimageio locally.

I will go with my previous comment that angle-grinder is probably the only formula affected by this update.
As rust is its only dependency, I don't expect us needing to rebuild the bottles.
If we need to fix, we can try modifying the problematic dependency that @Bo98 was looking into.

As more Rust projects are starting to use 1.52+ features, I think it is better to get this merged.

[carlocab]

Yes, let's merge this. It's blocking a few other PRs.

[BrewTestBot]

@carlocab has triggered a merge.