[BatchMode] Avoid spurious warnings in sourcekitd and indexing

[benlangmuir]

Explicitly disable batch mode in createCompilerInvocation, since it uses
-force-single-frontend-invocation. Previously we were getting spurious
warnings. Also add a test that -disable-batch-mode will allow commands
that use -index-file to avoid the same warning, since that is likely
what they want to do as well.

rdar://39581506

[benlangmuir]

@swift-ci please smoke test

[graydon]

Looks good to me but the last person to touch this logic was @davidungar so probably best to ask him.

[benlangmuir]

(new commit is to fix the commit message; no change to code)

[benlangmuir]

@swift-ci please smoke test

[davidungar]

Where is the -enable-batch-mode coming from? Would it be reasonable to remove that in addition to annihilating it with an anti-particle? Or is that not under our control in this circumstances? I like leaving the disable-batch-mode in for the day when batch mode becomes the default.

[davidungar]

How about adding a comment like the one on line 22, briefly describing the reason for the test? As a newbie, I've found such to help a lot.

[benlangmuir]

Sure, I'll add a follow-up commit explaining what my goal is here. Roughly:

The following test is verifying that -disable-batch-mode overrides an earlier -enable-batch-mode and silences the warning about mixing batch mode with -index-file. Tools that take an existing command line and add -index-file can add -disable-batch-mode without having to otherwise interpret the arguments.

[davidungar]

I've written a few thoughts down for your consideration, but none of them are critical.

[benlangmuir]

Where is the -enable-batch-mode coming from?

A build system or end user. Somewhere outside the compiler in any case.

Would it be reasonable to remove that in addition to annihilating it with an anti-particle? Or is that not under our control in this circumstances?

Do you mean remove it inside createCompileInvocation() or ask the end-user/build system to remove it? If you mean the former, then I guess we could although I don't see the advantage. If you mean the latter, then it is outside our control in the compiler and I don't think that external tools should need to care about how -enable-batch-mode interacts with implementation details of sourcekitd or indexing.