Looks like we have at least two build issues on OS X: 1. Bobby says OS X seems to still try to update the cargo registry from the network. Simon hit this locally on Linux. 2. -Wunreachable-code warnings-as-errors in Skia: https://treeherder.mozilla.org/logviewer.html#?job_id=26881763&repo=try gfx/skia/skia/src/core/SkBlitter_A8.cpp:382:10: warning: code will never be executed [-Wunreachable-code] 18:05:38 INFO - if(!(fDevice.width() == totalCount)) { do { SkNO_RETURN_HINT(); SkDebugf("%s:%d: fatal error: \"%s\"\n", "/builds/slave/try-m64-d-00000000000000000000/build/src/gfx/skia/skia/src/core/SkBlitter_A8.cpp", 382, "\"fDevice.width() == totalCount\""); ; sk_abort_no_print(); } while (false); }; This is surprising, however, because we build Skia with ALLOW_COMPILER_WARNINGS = True: https://hg.mozilla.org/mozilla-central/file/tip/gfx/skia/moz.build#l553

Also: 3. Linker assertion failures: https://treeherder.mozilla.org/logviewer.html#?job_id=26881764&repo=try#L22698 Assertion failed: (_mode == modeFinalAddress), function finalAddress, file /SourceCache/ld64/ld64-123.2.1/src/ld/ld.hpp, line 573. Nathan says: We use a custom-built linker: http://dxr.mozilla.org/mozilla-central/source/build/macosx/local-mozconfig.common#19 along with the custom LDFLAGS a bit further down in the same file; we did this to avoid link failures in libxul with newer Rust toolchains, but I guess we now have to pass those same flags to the Rust staticlib linking process as well? Is there reasonable support for that? (The linker hacks ought to be able to go away when we upgrade to Rust 1.12, as the unwind info problems have gone away, but until that point...)

(In reply to Chris Peterson [:cpeterson] from comment #0) > 2. -Wunreachable-code warnings-as-errors in Skia: > > https://treeherder.mozilla.org/logviewer.html#?job_id=26881763&repo=try > > gfx/skia/skia/src/core/SkBlitter_A8.cpp:382:10: warning: code will never be > executed [-Wunreachable-code] > 18:05:38 INFO - if(!(fDevice.width() == totalCount)) { do { > SkNO_RETURN_HINT(); SkDebugf("%s:%d: fatal error: \"%s\"\n", > "/builds/slave/try-m64-d-00000000000000000000/build/src/gfx/skia/skia/src/ > core/SkBlitter_A8.cpp", 382, "\"fDevice.width() == totalCount\""); ; > sk_abort_no_print(); } while (false); }; > > This is surprising, however, because we build Skia with > ALLOW_COMPILER_WARNINGS = True: > > https://hg.mozilla.org/mozilla-central/file/tip/gfx/skia/moz.build#l553 These are just ordinary warnings, not failures; for whatever reason, Treeherder is not smart enough to ignore them when figuring out where break-the-build errors are.

(In reply to Chris Peterson [:cpeterson] from comment #0) > 1. Bobby says OS X seems to still try to update the cargo registry from the > network. Simon hit this locally on Linux. Is this still a problem? Assuming stylo is using the stock Rust integration we have now, this shouldn't happen.

This isn't a problem anymore.