HLASM plugin does flag some long displacement (RXY) intructions as error

[cybersorcerer]

The instructions marked as bold are recognized as Syntax error by the Plugin. The error description is: Error at LAY instruction: value of address Operand displacement value must be between -2048 and 2047.
Example:

DoItAgain MVCOS 0(R14),0(R2),R10 Move data to output area
LAY R2,4096(,R2) Prepare source > 4 K
LAY R14,4096(,R14) Prepare target for > 4k
AHI R10,-4096 Check if bytes left
JP DoItAgain

There are two Problems with this.

1. The error would be correct if i had something like LA R2,4096 BUT the range in the error message i.e. -2048 and 2047 is wrong cause LA does not allow signed values. So if i had something like LA R2,-100(,R2) in my Code, the Plugin would be happy but HLASM would throw an "Invalid displacement" error, cause a 12 Bit displacement has to be between 0 and 4095.

2. The z990 processors support “long displacement (20-Bit)” RX-format instructions (RXY-a), with a displacement field range of -524288 to +524287 (± 0.5 Megabyte). All of the z/Architecture RX-format instructions introduced by the z900 are “long displacement” except for the floating point instructions. There are “Y” (long displacement) forms of most of the ESA/390 RX-format instructions except branch and as already mentioned floating Point intructions. Examples: AY, AHY, LY, LAY, STY

So the LAY instructions in the above example are perfectly valid and should never be flagged as error. For 12-Bit displacement intructions the error message text is wrong.

[michalbali256]

Thank you for nicely described bug!
I have already found it in code - unfortunately all 20 bit displacements were validated as if they were signed 12 bit displacements.

[cybersorcerer]

That was fast Michal, thanks! Found some other quirks today but i thought i collect them first and open one Report with what i have found so far. Did you also fix the wrong error message text? 12-Bit displacement have Always to be positiv numbers with a valid range of 0 to 4095.

[michalbali256]

Yes, the message is generated from the characteristic of the operand: number of bits and whether it is signed or unsigned.

[cybersorcerer]

Perfect!

[michalbali256]

The issue should be resolved by 0.11.0