RFR(XS): 8230669: [s390] C1: assert(is_bound() || is_unused()) failed: Label was never bound to a location, but it was used as a jmp target

[Liu, Xin]

Hi, Martin,

The intention is to catch malformed code in assembler. It also helps developers to write assembly code.
Indeed,  the assertion is kind of useless in bailout path. I sympathize your feeling.

your fix  is correct.  I found c1_LIRAssembler_s390.cpp has yet another place has similar pattern – line 2574.  Could you cover it as well?

BAILOUT cases scatter everywhere. I don’t want to see you get bitten again. I have a idea. How about we  group all local Labels and give them a shared state.
The state indicates that the current code path is in bailout or not.
Does it make sense?

Bailout state;
{

Label  a(state)
NearLabel b(state);
NearLabel c(state);

If (bailout) {
  state.set();
  // dtors of Label skip assert
  return ;
}

// dtors uses assert
}

Paul suggests to use overloaded macro for CHECK_BAILOUT.  It might bring compiler compatibility problem. Let me search hotspot code to see if it has supported overloaded and variadic macros.

Thanks,
--lx



From: "Doerr, Martin" <martin.doerr at sap.com>
Date: Thursday, September 5, 2019 at 9:57 AM
To: "Liu, Xin" <xxinliu at amazon.com>, "Hohensee, Paul" <hohensee at amazon.com>, "'hotspot-compiler-dev at openjdk.java.net'" <hotspot-compiler-dev at openjdk.java.net>
Subject: RFR(XS): 8230669: [s390] C1: assert(is_bound() || is_unused()) failed: Label was never bound to a location, but it was used as a jmp target

Hi,

the time bomb JDK-8206075<https://bugs.openjdk.java.net/browse/JDK-8206075> has hit us again (3rd time):
https://bugs.openjdk.java.net/browse/JDK-8230669

I think the recommended way to deal with it is:
http://cr.openjdk.java.net/~mdoerr/8230669_s390_label_asserts/webrev.00/

Please review.

Best regards,
Martin