Re: The Inbox: MethodMassage-dtl.71.mcz

This definitely needs expert review, I am a rookie in this area but if I 
am reading it right the fix is simple. If so, this gets us down to one 
remaining test failure in MessageMassageTests for both bytecode sets.

Dave

On 2026-05-10 20:49, [email protected] wrote:

> David T. Lewis uploaded a new version of MethodMassage to project The 
> Inbox:
> http://source.squeak.org/inbox/MethodMassage-dtl.71.mcz
> 
> ==================== Summary ====================
> 
> Name: MethodMassage-dtl.71
> Author: dtl
> Time: 10 May 2026, 4:49:49.186657 pm
> UUID: 06b25161-0bbf-4861-b500-11592dfae901
> Ancestors: MethodMassage-dtl.70
> 
> Fix for MessageMassageTest errors when running with 
> EncoderForV3PlusClosures.
> The errors are raised by an assert check in 
> ByteCodeAssembler>>spanToBlockLabel: which checks that the bytecode 
> preceding a block label is a return bytecode. Add 
> #methodReturnSpecialConstant to the list of possible returns.
> 
> =============== Diff against MethodMassage-dtl.70 ===============
> 
> Item was changed:
> ----- Method: BytecodeAssembler>>spanToBlockLabel: (in category 
> 'assembly') -----
> spanToBlockLabel: label "<Association>"
> "N.B. Block labels are expected to follow the last bytecode
> in the block and preceed the bytecode following the block."
> | labelIndex |
> labelIndex := label value.
> labelIndex ifNil:
> [labelIndex := code instructions identityIndexOf: label.
> label value: labelIndex].
> self assert: (labelIndex > 1
> and: [(self isLabel: (code instructions at: labelIndex - 1)) not
> +                 and: [#(blockReturnTop blockReturnConstant: 
> methodReturnSpecialConstant:
> -                 and: [#(blockReturnTop blockReturnConstant:
> methodReturnTop methodReturnConstant: methodReturnReceiver)
> includes: (code instructions at: labelIndex - 1) selector]]).
> ^(sizes at: labelIndex) ifNotNil:
> [self totalSizeFrom: index + 1 to: labelIndex - 1]!
> 
> Squeak-dev mailing list -- [email protected]
> To unsubscribe send an email to 
> [email protected]

Squeak-dev mailing list -- [email protected]
To unsubscribe send an email to [email protected]