Failure verify SOP¶

Mass failure can occur due to critical code change.
Or, if RD tried to introduce some fundamental enhance, which may break existing regression suite.

WARNING:¶

A COMPLETE FAILURE REPORT IS NECESSARY
Direct re-golden is prohibited

Procedure¶

Step 1: run regression and investigate log/reg.log
Step 2: try to understand what command failure and why
Step 3: if failure is caused by tiny variation, a correlation check is a must.

Discussion Demo:¶

Fix sanity check and re-golden related case. resistance_sanity_check_regolden.docx (ChunYan)
Feedback and regression failure sharing from Junc: supply_current_difference.pptx

See below examples:¶

1. PG mesh change, cause 60+ regression fail, each case has resistance and ir comparison to prove tiny variation. node_reduction_investigation.pptx (XiaoHu, DanPing)
2. twf, em code change cause 7+ regression fail, each case has log tcl investigation and solution pow_twf_em_fail_case.pptx (LangYun, Heng)
3. PG pad change, caused 20+ case fail, each fail needs detailed report and discussion use_pad_phypin_as_pg_with_feedback.pptx (Min)