tree b8457bca0fe20882d5495f661aa8cbe38bda8213
parent 9551980df2e46388565d0fea985eea6385d5c8e4
author Alon Zakai <azakai@google.com> 1772662857 -0800
committer GitHub <noreply@github.com> 1772662857 -0800
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsFcBAABCAAQBQJpqLBJCRC1aQ7uu5UhlAAA18YQAF7qFNGTVQedsYckPk136rWr
 fnhnMVMCDLLetWMPOBjBd7haOVBX3yOAUwmVX/YtGQKBOfOeA9zMGPaLVlJKsDtF
 0ymHjaKr8TIg0i7voVxkIOMBPdIHaPMg3LA6+qXXwH4iDujg50D/N5d/UCJ6XPX5
 l6i8IFjpsjKdRsVTNHkFqNwcOdSDiKkYjgyp8UcOqznFB/0NAOBNWLK91yvs+rH5
 u35jDl73Ty1aE7wOlfI5qVsfdU7hTfjaWBfXF7A5wRUrhVxpEXIPpEChvPD5UoBJ
 G+fQ20GGnSWXfuGuFBd1yV4bTuAOesC6idmG+ZrEo3p8Byk4/uFb+nT6J4M3TzDY
 kZ73xtlAZUEX1ijZij5lkVa+qZChlRqiJguKvNf93S7kTiHGJxwnR2rNl6EziBko
 uUJMc7V4+3TxCxsFctE88xMHNvsM+KFzLxEGxi+nmBdq43gqgJWLHQs5rRP3sZ/1
 4voRSUH5YoYZpV1JWrhvr9ZVi7qEP/5e3KKdgi65+Anrx3D0zI5BnI2nrm+vPZ26
 GpYeocjEJjqqsW1jzVsrovP+J9AJxuKfDyFHFwsGU0Jerwf6tt8GJSXQmrjUbDMD
 nmfObDC3ZQ2+Zgij+DblGrvwy11qQqTYLvFgEpOe5xQNGiOVz9/5vmlD4t6BaPwT
 jISiX8c32+2TXRO39nf6
 =iqJc
 -----END PGP SIGNATURE-----
 

OptimizeInstructions: Consecutive instrs are not equal if one tees and one reads (#8415)

`areConsecutiveInputsEqual` looks at fallthrough values when comparing,
but it must look at the original values for effects. It did that for the
right side, but not the left.

This was incorrect for a very long time, and only uncovered by

https://github.com/WebAssembly/binaryen/pull/8355#issuecomment-3976555920

I suspect it might have even been undiscoverable until then, even though
it was logically false (in that nothing reached it that could hit the bug).