I haven't used arc in a little while and am having trouble sorry for the emails

Arcanist accidentally threw away the first commit in the sequence...

Add changes suggested by Duncan to make CLANG_DEFAULT_STD_C more misuse-resistant

That looks good to me. I'm testing it out locally to make sure it still works right and also catches the issue, and will plan put up an updated patch in the morning.

Ah, so it'd be a test that passes pretty trivially on any bot that doesn't have a custom version of CLANG_DEFAULT_STD_C and CLANG_DEFAULT_STD_CXX like the default config, but could get caught by any other bots that do set it?

Unless anyone has an idea about how to test this, I'm going to commit this without adding a test.

I would be interested in adding some kind of test for this, but am not sure how.

This looks good to me.

Style nit on quote-vs-slash, but otherwise this looks good to me.

ping on this one, since it's the first in the stack and blocking the others

Flip argument order of EXPECT_EQ for better failure messages

Flip argument order of EXPECT_EQ for better failure messages

Flip argument order of EXPECT_EQ for better failure messages

Change some EXPECT_TRUE to EXPECT_EQ

Change some EXPECT_TRUE to EXPECT_EQ

Change some ASSERT_TRUE to ASSERT_EQ

Change Optional<Register> to Register, change a >= to == since it's all that's needed

Remove ";" file I accidentally added

Change unittest buildInstr to more specific build* calls

Change unittest buildInstr to buildSAddo/buildSSubo

Add comment about number of defs, change buildInstr

Change Optional<Register> to Register

I can commit them.

The MIR syntax is based on extending the yaml syntax and then embedding the machine-ir syntax inside it, I've defined it as two separate files so it's easier to separate "highlighting the instructions" from "embedding it inside yaml".

ping, are there any changes this still needs? I think the existing comments are sufficiently addressed given that the goal is to make this an NFC change and fix them up separately.

Use a temporary variable for the destination register/type

Re-add the vector code to make it more properly an NFC commit

Moved unreachable case to bottom of switch

Add a default case assertion, mark vectors as unsupported, cleanups

A question for reviewers who might be able to explain:

Add unreachable case, give vregs names in tests

Fix broken named vreg in legalize-ssube.mir

Used named vregs in test cases

Add unit-tests in LegalizerHelperTest

Change MachineOperand variable to Register, use getReg(0) on MachineInstrBuilder

Update tests to match run-line changes

Update test run line to match suggestions on D95323

Update test run line to match suggestions on D95323

Change the run-line for the new test files

The buildbot with the UBSan tests passed for the first commit in this stack, so I think this should be safe to re-land http://lab.llvm.org:8011/#/builders/77/builds/3112

The buildbot with the UBSan tests passed for the first commit in this stack, so I think this should be safe to re-land http://lab.llvm.org:8011/#/builders/77/builds/3112

I also reopened D95032 with the fix squashed onto that, please review whichever of them will be more convenient.

Fix instruction selection for G_USUBO