Don't truncate compiler output in compiletest

[ashtneoi]

Fixes issue #53332. Should I write a test for this, or is it sufficiently regression-proof?

[rust-highfive]

r? @alexcrichton

(rust_highfive has picked a reviewer for you, use r? to override)

[alexcrichton]

Thanks for the PR! I want to be sure to cc @kennytm in relation to #45783 where this was originally added. It seems that if we do this we may go back to OOMing sometimes?

A better default may be to perhaps only whitelist tests to be executed with read2_abbreviated?

[kennytm]

read2_abbreviated is definitely needed as long as the test run-pass\rustc-rust-log.rs still runs on a 32-bit platform.

Perhaps enable read2_abbreviated just for run-pass tests?

P.S. Why would the JSON output exceed 416 KB? Could we redact the error explanation?

[ashtneoi]

read2_abbreviated is definitely needed as long as the test run-pass\rustc-rust-log.rs still runs on a 32-bit platform.

Aha, I knew I was missing something.

P.S. Why would the JSON output exceed 416 KB? Could we redact the error explanation?

That sounds like a much better idea. Maybe we could add a new --error-format option that omits explanations? Call it something like json-brief or json-no-explanations?

[alexcrichton]

It's probably best to avoid changing the compiler here, ideally we'd just use the abbreviated reading for that one test I believe.

[pietroalbini]

Ping from triage! What's the status of this PR?

[ashtneoi]

I think we're waiting on me now, but I didn't have much time last week to work on it (sorry).

[ashtneoi]

I'm not sure when I'll get time to finish this so I'm going to close it for now. If someone else wants to work on it, they're welcome to.