/azp run

/azp run

Hi @kewei-arista, thanks for the detailed fix!

I noticed this PR overlaps with #22829 (longhuan-cisco) — both address the same root issue: stale/historical FEC uncorrectable errors causing test_verify_fec_stats_counters to falsely fail.

A few issues I found in this PR:

Bug 1: NameError in except ValueError handlers (2 locations)

In get_interface_snapshots and test_verify_fec_stats_counters, the except ValueError block references the variable that failed to be assigned:

try:
    fec_uncorr_int = int(fec_uncorr)
except ValueError:
    pytest.fail("... fec_uncorr: {}".format(intf_name, fec_uncorr_int))  # NameError!

Should use fec_uncorr (the raw string) instead of fec_uncorr_int.

Bug 2: Global state never cleared — breaks multi-HWSKU testbeds

INTF_STATUS_SNAPSHOTS and INTF_FEC_HISTOGRAM_SNAPSHOTS are module-level lists that are never cleared. With enum_rand_one_per_hwsku_frontend_hostname, on a multi-HWSKU testbed:

• Run 1 (dut1) → populates globals
• Run 2 (dut2) → if not INTF_STATUS_SNAPSHOTS is False → silently uses dut1's data to validate dut2

This silently validates the wrong DUT's counters.

Design concern: Cross-test state sharing

The shared global mutable state between test_verify_fec_stats_counters and test_verify_fec_histogram makes tests order-dependent and fragile. Consider using a module-scoped pytest fixture instead.

Comparison with #22829:

PR #22829 takes a simpler approach (clear counters + wait 60s) that avoids these issues. However, this PR has the advantage of also optimizing the histogram test polling. Perhaps the histogram optimization could be a separate follow-up PR once the bugs above are addressed?

@kewei-arista can we use this PR? #22829