awk Part III · Module 9 · Simulation and UVM Log Analysis

Counting Warnings and Errors

A triage reads two or three numbers before anything else: how many errors, how many warnings, did anything fatal happen. Producing them looks like the simplest job in this module and it contains the module's sharpest trap — the obvious count is wrong, and the reflex fix of anchoring the pattern does not help, because the log's own report summary begins with the same word you are matching. Past that, a harder question: the number you have just computed counts message records, and a triage wants failures. On the fixture here, nine error records are two distinct failures.

Intermediate16 min readawkPart IIIUVMTriageCounting

Part III · Module 9 · Chapter 9.3 · Counting Warnings and Errors

1. Why This Matters

This is the most-written awk command in verification: count the errors in a log. It appears in every regression script, usually as a grep -c, and it is usually wrong.

It is rated critical because of where the number goes. An error count decides whether a nightly job reports success, whether a test is retried, and whether anybody looks at the log at all. A count that is high by one makes a clean run look dirty; a count that is low by one makes a failing run pass.

2. The Obvious Count

Azvya Education Pvt. Ltd.VLSI Mentor

sim.log — synthetic teaching data from Chapter 9.1. Modelled on the UVM default report format.

Chronologic VCS simulator copyright 1991-2024
Compiler version X-2024.03; Runtime version X-2024.03;  Oct 12 14:30 2026
UVM_INFO @ 0: reporter [RNTST] Running test alu_smoke_test...
UVM_INFO ../tb/env/alu_env.sv(87) @ 0: uvm_test_top.env [BUILD] env built with 2 agents
UVM_INFO ../tb/seq/alu_seq.sv(42) @ 1000: uvm_test_top.env.agt.seqr@@seq [SEQ] starting 50 items
UVM_WARNING ../tb/env/alu_sb.sv(210) @ 15500: uvm_test_top.env.sb [UNUSED_CLK] clock gate never toggled
UVM_WARNING ../tb/env/alu_sb.sv(210) @ 16000: uvm_test_top.env.sb [UNUSED_CLK] clock gate never toggled
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 20500: uvm_test_top.env.sb [MISMATCH] exp=32'hdead act=32'hbeef
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 21000: uvm_test_top.env.sb [MISMATCH] exp=32'hcafe act=32'h0000
UVM_INFO ../tb/seq/alu_seq.sv(99) @ 30000: uvm_test_top.env.agt.seqr@@seq [SEQ] done
UVM_FATAL ../tb/env/alu_env.sv(300) @ 30500: uvm_test_top.env [DEADLOCK] no response for 500ns
--- UVM Report Summary ---
** Report counts by severity
UVM_INFO :    4
UVM_WARNING :    2
UVM_ERROR :    2
UVM_FATAL :    1
$finish called from file "../tb/top.sv", line 55.

The log has two error records. Count them:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

grep -c UVM_ERROR sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/UVM_INFO/ { i++ } /UVM_WARNING/ { w++ } /UVM_ERROR/ { e++ } /UVM_FATAL/ { f++ } END { printf "INFO=%d WARNING=%d ERROR=%d FATAL=%d  (of %d lines)\n", i + 0, w + 0, e + 0, f + 0, NR }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

INFO=5 WARNING=3 ERROR=3 FATAL=2  (of 18 lines)

Every count is high by one, and the log itself says so three lines from the end: UVM_INFO : 4, UVM_WARNING : 2, UVM_ERROR : 2, UVM_FATAL : 1.

3. Why, and Why Anchoring Does Not Help

Printing what matched settles the cause:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/UVM_ERROR/ { printf "line %d: %s\n", NR, $0 }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

line 8: UVM_ERROR ../tb/env/alu_sb.sv(145) @ 20500: uvm_test_top.env.sb [MISMATCH] exp=32'hdead act=32'hbeef
line 9: UVM_ERROR ../tb/env/alu_sb.sv(145) @ 21000: uvm_test_top.env.sb [MISMATCH] exp=32'hcafe act=32'h0000
line 16: UVM_ERROR :    2

Line 16 is from the report summary — the log's own count of errors, matched as if it were an error.

The reflex fix is to anchor the pattern to the start of the line. It does not work:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^UVM_INFO/ { i++ } /^UVM_WARNING/ { w++ } /^UVM_ERROR/ { e++ } /^UVM_FATAL/ { f++ } END { printf "INFO=%d WARNING=%d ERROR=%d FATAL=%d\n", i + 0, w + 0, e + 0, f + 0 }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

INFO=5 WARNING=3 ERROR=3 FATAL=2

Identical. UVM_ERROR : 2 also starts with UVM_ERROR, so the anchor excludes nothing.

The fix that feels right does nothing, which is why this ships

Anchoring is the standard remedy for an over-matching pattern and it is the first thing an experienced engineer tries. Here it changes no number at all.

So the sequence that puts this bug into production is: write grep -c UVM_ERROR, get a count that seems plausible, and never compare it against anything. Or: notice the count is one too high, anchor the pattern, observe that the command still runs, and move on without re-checking the number.

The second path is worse, because it leaves a comment or a commit message claiming the problem was fixed.

What makes the summary line so hard to exclude is that it is deliberately similar — UVM prints its own severity names in its own summary, so any pattern built from the severity name matches both a record and the claim about the records. Nothing short of using more of the line's structure will separate them.

Section 4 gives three ways. The one worth internalising is the third: describe what a record looks like, rather than listing what to exclude. Chapter 5.3 section 11 argued the same thing for numeric validation, and Chapter 7.6 section 7 for CSV detection — a negative pattern is only as complete as your imagination, and here the thing you failed to imagine is in the file every time.

4. Three Fixes

Skip the summary region. The structural answer — once past the marker, nothing is a record:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1 } insum { next } /^UVM_INFO/ { i++ } /^UVM_WARNING/ { w++ } /^UVM_ERROR/ { e++ } /^UVM_FATAL/ { f++ } END { printf "INFO=%d WARNING=%d ERROR=%d FATAL=%d\n", i + 0, w + 0, e + 0, f + 0 }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

INFO=4 WARNING=2 ERROR=2 FATAL=1

Require the second field not to be a colon. The summary's layout is SEVERITY : N, so $2 is : there and a path or @ in a record:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '$2 != ":" && /^UVM_INFO/ { i++ } $2 != ":" && /^UVM_WARNING/ { w++ } $2 != ":" && /^UVM_ERROR/ { e++ } $2 != ":" && /^UVM_FATAL/ { f++ } END { printf "INFO=%d WARNING=%d ERROR=%d FATAL=%d\n", i + 0, w + 0, e + 0, f + 0 }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

INFO=4 WARNING=2 ERROR=2 FATAL=1

A positive record pattern. Describe a record: a severity, a space, then a path or an at-sign:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^UVM_(INFO|WARNING|ERROR|FATAL) (\.|\/|@)/ { n[$1]++ } END { printf "INFO=%d WARNING=%d ERROR=%d FATAL=%d\n", n["UVM_INFO"] + 0, n["UVM_WARNING"] + 0, n["UVM_ERROR"] + 0, n["UVM_FATAL"] + 0 }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

INFO=4 WARNING=2 ERROR=2 FATAL=1

All three correct. They differ in what they assume:

Fix	Assumes	Breaks if
skip the summary	the marker text	a tool uses different wording
`$2 != ":"`	the summary's spacing	the summary is reformatted
positive pattern	a record's own shape	a record format is added that it does not cover

The third is used throughout Module 9, and its failure mode is the mildest — it under-counts rather than over-counts, and Chapter 9.1 section 6's cross-check catches that.

5. Checking the Count Against the Log

With the count correct, verify it rather than trusting it:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1; next } insum && /^UVM_[A-Z]+ *:/ { line = $0; gsub(/ /, "", line); split(line, p, ":"); claimed[p[1]] = p[2] } !insum && /^UVM_(INFO|WARNING|ERROR|FATAL) (\.|\/|@)/ { counted[$1]++ } END { bad = 0; for (s in claimed) { ok = (counted[s] + 0 == claimed[s] + 0); printf "%-12s counted=%d  log=%s  %s\n", s, counted[s] + 0, claimed[s], (ok ? "agree" : "DISAGREE"); if (!ok) bad++ } exit (bad ? 1 : 0) }' sim.log | sort

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

UVM_ERROR    counted=2  log=2  agree
UVM_FATAL    counted=1  log=1  agree
UVM_INFO     counted=4  log=4  agree
UVM_WARNING  counted=2  log=2  agree

Four for four. The exit status makes it a gate: a disagreement means the pattern missed a record format, and the script should stop rather than report a number it cannot defend.

The naive version fails that gate immediately:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1; next } insum && /^UVM_[A-Z]+ *:/ { line = $0; gsub(/ /, "", line); split(line, p, ":"); claimed[p[1]] = p[2]; next } /^UVM_[A-Z]+/ { counted[$1]++ } END { for (s in claimed) printf "%-12s counted=%d  log=%s  %s\n", s, counted[s] + 0, claimed[s], (counted[s] + 0 == claimed[s] + 0 ? "agree" : "DISAGREE") }' sim.log | sort

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

UVM_ERROR    counted=2  log=2  agree
UVM_FATAL    counted=1  log=1  agree
UVM_INFO     counted=4  log=4  agree
UVM_WARNING  counted=2  log=2  agree

That one agrees too — because consuming the summary lines with next happens to remove exactly the lines that were causing the over-count. It is correct by accident: the next was written to stop summary lines reaching the counter, which is fix A in a different spelling.

6. The Triage Numbers

What a triage actually wants is three numbers and a verdict:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1 } insum { next } /^UVM_(INFO|WARNING|ERROR|FATAL) (\.|\/|@)/ { n[$1]++ } END { e = n["UVM_ERROR"] + 0; f = n["UVM_FATAL"] + 0; w = n["UVM_WARNING"] + 0; printf "%d error(s), %d fatal, %d warning(s) -> %s\n", e, f, w, (f ? "FATAL" : (e ? "FAIL" : "PASS")) }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

2 error(s), 1 fatal, 2 warning(s) -> FATAL

The verdict order matters and is an engineering decision, not a formatting one. A UVM_FATAL means the simulation stopped — so whatever else the log contains, it is incomplete, and treating the run as merely "failing" understates it. Chapter 9.5 makes FATAL a distinct status for exactly this reason.

Warnings are reported and do not affect the verdict here, which is the common convention and worth stating explicitly rather than leaving implicit — a project that gates on warnings needs a different rule, and a reader should be able to see which this script implements.

7. A Count of Records Is Not a Count of Failures

The second half of the chapter, and the harder half. A log with many errors:

Azvya Education Pvt. Ltd.VLSI Mentor

logs/alu.log — synthetic. Nine error records from two distinct causes.

UVM_INFO @ 0: reporter [RNTST] Running test alu_test...
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1000: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1100: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1200: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1300: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1400: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1500: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1600: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1700: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_cov.sv(77) @ 2000: uvm_test_top.env.cov [NO_HIT] bin never hit
--- UVM Report Summary ---
UVM_ERROR :    9

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1 } insum { next } /^UVM_ERROR (\.|\/|@)/ { e++ } END { printf "%d error record(s)\n", e + 0 }' logs/alu.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

9 error record(s)

Nine, agreeing with the log. Now count distinct failures, keying on the id and the source location:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1 } insum { next } /^UVM_ERROR (\.|\/|@)/ { recs++; id = $0; sub(/^[^[]*\[/, "", id); sub(/\].*$/, "", id); sig[id SUBSEP $2]++ } END { c = 0; for (k in sig) c++; printf "%d error record(s) collapse to %d distinct signature(s)\n", recs, c }' logs/alu.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

9 error record(s) collapse to 2 distinct signature(s)

Nine records, two signatures. Both numbers are true and they answer different questions:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1 } insum { next } /^UVM_ERROR (\.|\/|@)/ { id = $0; sub(/^[^[]*\[/, "", id); sub(/\].*$/, "", id); loc = $2; sub(/\(/, ":", loc); sub(/\)/, "", loc); sig[id SUBSEP loc]++ } END { for (k in sig) { split(k, p, SUBSEP); printf "%-12s %-28s %d record(s)\n", p[1], p[2], sig[k] } }' logs/alu.log | sort -k3 -rn

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

MISMATCH     ../tb/env/alu_sb.sv:145      8 record(s)
NO_HIT       ../tb/env/alu_cov.sv:77      1 record(s)

Nine is how loud the failure is; two is how many things are wrong

This is Chapter 6.2 section 5's records-are-not-things, in the setting where it costs the most.

The eight MISMATCH records are almost certainly one bug — a scoreboard comparing every transaction, reporting each mismatch, on a design that is wrong in one way. The one NO_HIT is a second, different problem, and it is the row a triage would otherwise never see, because it is one line in nine.

Which means the two numbers support opposite decisions. 9 errors says this test failed badly and somebody should spend a day on it. 2 distinct failures says there are two things to fix, one of which is a coverage hole that may be unrelated to the other.

And the ratio is not stable. The same single bug produces 8 records on a 50-transaction test and 800 on a 5,000-transaction one, so the record count partly measures the test length rather than the design. Comparing error counts between two runs of different lengths is comparing the stimulus, not the quality.

So the rule for a triage table: report both, and sort by the distinct count. Chapter 9.4 builds that table properly. The record count still matters — it tells you how reproducible a failure is — but it is the wrong number to rank by, exactly as Chapter 6.2 section 6 argued about frequency.

8. Both Numbers, Side by Side

The command a regression script should actually run:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1 } insum { next } /^UVM_(ERROR|FATAL) (\.|\/|@)/ { recs++; id = $0; sub(/^[^[]*\[/, "", id); sub(/\].*$/, "", id); sig[id SUBSEP $2]++; if ($1 == "UVM_FATAL") fatal++ } END { c = 0; for (k in sig) c++; printf "%d record(s), %d distinct signature(s), %d fatal -> %s\n", recs + 0, c, fatal + 0, (fatal ? "FATAL" : (recs ? "FAIL" : "PASS")) }' logs/alu.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

9 record(s), 2 distinct signature(s), 0 fatal -> FAIL

And on the Chapter 9.1 fixture, where the counts are small and the fatal is present:

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1 } insum { next } /^UVM_(ERROR|FATAL) (\.|\/|@)/ { recs++; id = $0; sub(/^[^[]*\[/, "", id); sub(/\].*$/, "", id); sig[id SUBSEP $2]++; if ($1 == "UVM_FATAL") fatal++ } END { c = 0; for (k in sig) c++; printf "%d record(s), %d distinct signature(s), %d fatal -> %s\n", recs + 0, c, fatal + 0, (fatal ? "FATAL" : (recs ? "FAIL" : "PASS")) }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

3 record(s), 2 distinct signature(s), 1 fatal -> FATAL

Three records, two signatures, one fatal. The MISMATCH pair collapses to one signature and the DEADLOCK is the other.

Note the signature key uses $2 — the raw file(line) — rather than the id alone. Two checkers can share an id, and the same checker at two lines is usually two different checks, so the pair is the more honest grouping. Chapter 9.4 examines that choice.

9. Common Mistakes

grep -c UVM_ERROR. The report summary matches — section 2.

Anchoring and assuming it is fixed. The summary line also starts with the severity — section 3.

A negative pattern listing exceptions. Only as complete as your imagination.

Not checking against the log's own summary. It is a free independent count — section 5.

Treating a record count as a failure count. Nine records, two failures — section 7.

Comparing error counts between runs of different length. The count partly measures the test length.

Ranking a triage table by record count. The loudest failure is rarely the most important — Chapter 6.2 §6.

Reporting FAIL when a fatal occurred. The run stopped, so the log is incomplete — section 6.

Leaving the warning policy implicit. State whether warnings gate.

10. Exercises

Exercise 1 — Reproduce the over-count

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/UVM_ERROR/ { e++ } END { printf "naive=%d\n", e + 0 }' sim.log
awk '/^UVM_ERROR/ { e++ } END { printf "anchored=%d\n", e + 0 }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

naive=3
anchored=3

Both wrong, and identically so.

Exercise 2 — Find the extra match

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/UVM_ERROR/ { printf "line %d: %s\n", NR, $0 }' sim.log | tail -1

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

line 16: UVM_ERROR :    2

The log's own error count, counted as an error.

Exercise 3 — Count correctly

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^UVM_(INFO|WARNING|ERROR|FATAL) (\.|\/|@)/ { n[$1]++ } END { for (s in n) printf "%-12s %d\n", s, n[s] | "sort" }' sim.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

UVM_ERROR    2
UVM_FATAL    1
UVM_INFO     4
UVM_WARNING  2

Exercise 4 — Collapse records to signatures

Azvya Education Pvt. Ltd.VLSI Mentor

logs/alu.log

UVM_INFO @ 0: reporter [RNTST] Running test alu_test...
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1000: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1100: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1200: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1300: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1400: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1500: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1600: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_sb.sv(145) @ 1700: uvm_test_top.env.sb [MISMATCH] exp!=act
UVM_ERROR ../tb/env/alu_cov.sv(77) @ 2000: uvm_test_top.env.cov [NO_HIT] bin never hit
--- UVM Report Summary ---
UVM_ERROR :    9

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '/^--- UVM Report Summary ---/ { insum = 1 } insum { next } /^UVM_ERROR (\.|\/|@)/ { recs++; id = $0; sub(/^[^[]*\[/, "", id); sub(/\].*$/, "", id); sig[id SUBSEP $2]++ } END { c = 0; for (k in sig) c++; printf "%d record(s) -> %d signature(s)\n", recs, c }' logs/alu.log

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

9 record(s) -> 2 signature(s)

Exercise 5 — Engineering application

Produce a triage header for a log: both counts, the fatal flag, a verdict, and a cross-check against the log's own summary with a non-zero status if they disagree.

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk '
/^--- UVM Report Summary ---/ { insum = 1 }
insum && /^UVM_[A-Z]+ *:/ { line = $0; gsub(/ /, "", line); split(line, p, ":"); claimed[p[1]] = p[2] }
insum { next }
/^UVM_(INFO|WARNING|ERROR|FATAL) (\.|\/|@)/ {
    counted[$1]++
    if ($1 == "UVM_ERROR" || $1 == "UVM_FATAL") {
        id = $0; sub(/^[^[]*\[/, "", id); sub(/\].*$/, "", id)
        sig[id SUBSEP $2]++
    }
}
END {
    if (!NR) { print "ERROR: empty log" > "/dev/stderr"; exit 2 }
    e = counted["UVM_ERROR"] + 0; f = counted["UVM_FATAL"] + 0; w = counted["UVM_WARNING"] + 0
    c = 0; for (k in sig) c++
    printf "%d error(s) + %d fatal in %d distinct signature(s); %d warning(s) -> %s\n", e, f, c, w, (f ? "FATAL" : (e ? "FAIL" : "PASS"))
    bad = 0
    for (s in claimed) if (counted[s] + 0 != claimed[s] + 0) { printf "CROSS-CHECK FAILED %s: parsed %d, log says %s\n", s, counted[s] + 0, claimed[s] > "/dev/stderr"; bad++ }
    if (bad) exit 2
    if (f || e) exit 1
}' sim.log
echo "status: $?"

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

2 error(s) + 1 fatal in 2 distinct signature(s); 2 warning(s) -> FATAL
status: 1

Everything a triage needs in one line, with the record count and the signature count both present so neither can be mistaken for the other.

The status convention is Chapter 8.3 section 10's: 1 for a run that failed, 2 for a log that could not be trusted. A cross-check failure produces 2 rather than 1 — because if the parser disagrees with the log, the verdict it just printed is not reliable either.

The engineering caveat: 2 distinct signature(s) groups on (id, file(line)), which treats two checkers with the same id at different lines as different failures and the same checker firing on different data as one. Both are defensible readings and Chapter 9.4 is where the choice gets examined rather than assumed.

Exercise 6 — Challenge: show the record count tracking the test length

The claim in section 7 is that a record count partly measures the stimulus. Demonstrate it.

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

awk 'BEGIN { print "UVM_INFO @ 0: reporter [RNTST] Running test short..." ; for (i = 1; i <= 3; i++) printf "UVM_ERROR ../tb/env/alu_sb.sv(145) @ %d: uvm_test_top.env.sb [MISMATCH] exp!=act\n", i * 100 }' > short.log
awk 'BEGIN { print "UVM_INFO @ 0: reporter [RNTST] Running test long..." ; for (i = 1; i <= 40; i++) printf "UVM_ERROR ../tb/env/alu_sb.sv(145) @ %d: uvm_test_top.env.sb [MISMATCH] exp!=act\n", i * 100 }' > long.log
for f in short.log long.log; do
  printf "%-10s " "$f"
  awk '/^UVM_ERROR (\.|\/|@)/ { recs++; id = $0; sub(/^[^[]*\[/, "", id); sub(/\].*$/, "", id); sig[id SUBSEP $2]++ } END { c = 0; for (k in sig) c++; printf "%3d record(s), %d signature(s)\n", recs + 0, c }' "$f"
done

Azvya Education Pvt. Ltd.VLSI Mentor

Snippet

short.log    3 record(s), 1 signature(s)
long.log    40 record(s), 1 signature(s)

The record count differs by a factor of 13 and the signature count is identical. Both logs contain the same single bug reported by the same checker; the only difference is how many transactions the test ran before giving up.

So a dashboard tracking error counts over time would show this bug getting thirteen times worse when somebody lengthened the test. A dashboard tracking distinct signatures would show no change — correctly, because nothing about the design changed.

That is the argument for the signature count in one measurement. The record count is not useless: 40 records from 40 transactions says the failure is fully reproducible, where 1 record in 40 would say it is intermittent, and that distinction matters for debugging. It is simply not a measure of how much is wrong.

11. Summary

grep -c UVM_ERROR over-counts, because the log's report summary contains the severity name.
Anchoring the pattern does not fix it — the summary line starts with the severity too.
Three fixes work: skip the summary region, require $2 != ":", or use a positive record pattern. Prefer the positive pattern.
Cross-check the count against the log's own summary and exit non-zero on a disagreement.
A UVM_FATAL means the run stopped, so it deserves a status distinct from FAIL.
State the warning policy rather than leaving it implicit.
A record count is not a failure count. Nine records collapsed to two signatures here.
The record count partly measures the test length — the same bug gave 3 records and 40 on different-length runs, with 1 signature both times.
Report both numbers, and rank a triage by the signature count.
The record count still tells you reproducibility, which is why it is worth keeping.

Two numbers now exist for every log. Chapter 9.4 — Grouping Errors for Triage builds the table those numbers summarise: errors grouped by id and source so a hundred identical messages become one row, with the question of which key to group on examined rather than assumed.

Where this fits

Part of the awk curriculum.