nm01: Fix BSD vs POSIX format comparison failure on Power
PATCH 1/1 nm01: Fix BSD vs POSIX format comparison failure on Power
On Power architecture, symbol names can contain hex-like prefixes such as "0000003e.plt_call.__gmon_start__".
In our openqa specific power system, the gensub seems has some compatible issue which you can see following trace, the BSD format's symbol name was incorrectly trimmed to ".plt_call.__gmon_start__" (losing the "3e"), while the POSIX format kept it as "3e.plt_call.__gmon_start__".
Example failure from a Power machine:
--- DEBUG nm01_sh ---
--- RAW BSD ---00000000000005e0 t 0000003e.plt_call.__gmon_start__ --- RAW POSIX --- 0000003e.plt_call.__gmon_start__ t 5e0
--- REPRO LTP LOGIC ---
--- TRIMMED BSD ---
5e0 t .plt_call.__gmon_start__
--- TRIMMED POSIX --- 3e.plt_call.__gmon_start__ t 5e0
--- REORDERED BSD (nm1) --- .plt_call.__gmon_start__t5e0
--- REORDERED POSIX (nm2) --- 3e.plt_call.__gmon_start__t5e0
--- DIFF RESULT ---
tmp/dbg_nm2 renamed from tmp/dbg_nm1 +1 -2
@@ -1,7 +1,7 @@ -.plt_call.__gmon_start__t5e0 +3e.plt_call.__gmon_start__t5e0 --- END DEBUG ---
Replace the global gensub to simple sub approach with targeted field processing in awk can fix the issue.
--- testcases/commands/nm/nm01.sh | 21 ++++++++++++++++----- 1 file changed, 16 insertions(+), 5 deletions(-)
testcases/commands/nm/nm01.sh +16 -5
@@ -83,17 +83,28 @@ test5() EXPECT_PASS $NM -f bsd $TST_DATAROOT/f1 \> nm_bsd.out EXPECT_PASS $NM -f posix $TST_DATAROOT/f1 \> nm_posix.out - ROD awk '{print gensub(/\y(0+)([0-9a-fA-F]+)\y/, "\\2", "g")}' nm_bsd.out \> trimmed_nm_bsd.out - ROD awk '{print gensub(/\y(0+)([0-9a-fA-F]+)\y/, "\\2", "g")}' nm_posix.out \> trimmed_nm_posix.out - - ROD awk '{print $3 $2 $1}' trimmed_nm_bsd.out \> nm1.out - ROD awk '{print $1 $2 $3}' trimmed_nm_posix.out \> nm2.out + ROD awk '{ + if (NF == 2) { val = "0"; type = $1; name = $2; } + else { val = $1; type = $2; name = $3; } + sub(/^0+/, "", val); if (val == "") val = "0"; + sub(/^0+/, "", name); + print name type val + }' nm_bsd.out \> nm1.out + + ROD awk '{ + val = $3; type = $2; name = $1; + sub(/^0+/, "", val); if (val == "") val = "0"; + sub(/^0+/, "", name); + print name type val + }' nm_posix.out \> nm2.out if diff nm1.out nm2.out > /dev/null; then tst_res TPASS "Got BSD format with -f bsd" else tst_res TFAIL "Got wrong format with -f bsd" cat nm_bsd.out + cat nm_posix.out + diff -u nm1.out nm2.out fi }
Signature · 5 lines
2.52.0 -- Mailing list info: https://lists.linux.it/listinfo/ltp
Hi Wei,
patch is not applying
https://patchwork.ozlabs.org/project/ltp/patch/[email protected]/
--
Andrea Cervesato
SUSE QE Automation Engineer Linux
[email protected]
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
PATCH v2 1/1 nm01: Fix BSD vs POSIX format comparison failure on Power
In reply to Wei Gao via ltpOn Power architecture, symbol names can contain hex-like prefixes such as "0000003e.plt_call.__gmon_start__".
In our openqa specific power system, the gensub seems has some compatible issue which you can see following trace, the BSD format's symbol name was incorrectly trimmed to ".plt_call.__gmon_start__" (losing the "3e"), while the POSIX format kept it as "3e.plt_call.__gmon_start__".
Example failure from a Power machine:
--- DEBUG nm01_sh ---
--- RAW BSD ---
00000000000005e0 t 0000003e.plt_call.__gmon_start__ --- RAW POSIX --- 0000003e.plt_call.__gmon_start__ t 5e0
--- REPRO LTP LOGIC ---
--- TRIMMED BSD ---
5e0 t .plt_call.__gmon_start__
--- TRIMMED POSIX --- 3e.plt_call.__gmon_start__ t 5e0
--- REORDERED BSD (nm1) --- .plt_call.__gmon_start__t5e0
--- REORDERED POSIX (nm2) --- 3e.plt_call.__gmon_start__t5e0
--- DIFF RESULT ---
--- /tmp/dbg_nm1 2026-05-11 22:33:54.754648941 -0400
+++ /tmp/dbg_nm2 2026-05-11 22:33:54.761924742 -0400
@@ -1,7 +1,7 @@ -.plt_call.__gmon_start__t5e0 +3e.plt_call.__gmon_start__t5e0
--- END DEBUG ---
Replace the global gensub to simple sub approach with targeted field processing in awk can fix the issue.
---v1->v2:
Update commit message to fix patch apply error
testcases/commands/nm/nm01.sh | 21 ++++++++++++++++----- 1 file changed, 16 insertions(+), 5 deletions(-)
testcases/commands/nm/nm01.sh +16 -5
@@ -83,17 +83,28 @@ test5() EXPECT_PASS $NM -f bsd $TST_DATAROOT/f1 \> nm_bsd.out EXPECT_PASS $NM -f posix $TST_DATAROOT/f1 \> nm_posix.out - ROD awk '{print gensub(/\y(0+)([0-9a-fA-F]+)\y/, "\\2", "g")}' nm_bsd.out \> trimmed_nm_bsd.out - ROD awk '{print gensub(/\y(0+)([0-9a-fA-F]+)\y/, "\\2", "g")}' nm_posix.out \> trimmed_nm_posix.out - - ROD awk '{print $3 $2 $1}' trimmed_nm_bsd.out \> nm1.out - ROD awk '{print $1 $2 $3}' trimmed_nm_posix.out \> nm2.out + ROD awk '{ + if (NF == 2) { val = "0"; type = $1; name = $2; } + else { val = $1; type = $2; name = $3; } + sub(/^0+/, "", val); if (val == "") val = "0"; + sub(/^0+/, "", name); + print name type val + }' nm_bsd.out \> nm1.out + + ROD awk '{ + val = $3; type = $2; name = $1; + sub(/^0+/, "", val); if (val == "") val = "0"; + sub(/^0+/, "", name); + print name type val + }' nm_posix.out \> nm2.out if diff nm1.out nm2.out > /dev/null; then tst_res TPASS "Got BSD format with -f bsd" else tst_res TFAIL "Got wrong format with -f bsd" cat nm_bsd.out + cat nm_posix.out + diff -u nm1.out nm2.out fi }
Signature · 5 lines
2.52.0 -- Mailing list info: https://lists.linux.it/listinfo/ltp
Hi Wei,
testcases/commands/nm/nm01.sh test5() · line 92
Quoted 10 lines · Wei Gao
nm01: Fix BSD vs POSIX format comparison failure on Power
testcases/commands/nm/nm01.sh +7
-1- ROD awk '{print gensub(/\y(0+)([0-9a-fA-F]+)\y/, "\\2", "g")}' nm_bsd.out \> trimmed_nm_bsd.out + ROD awk '{ + if (NF == 2) { val = "0"; type = $1; name = $2; } + else { val = $1; type = $2; name = $3; } + sub(/^0+/, "", val); if (val == "") val = "0"; + sub(/^0+/, "", name); + print name type val + }' nm_bsd.out \> nm1.out
Good fix. Replacing gensub() (gawk extension) with sub() is the right approach here and correctly handles the hex-prefix symbol names on Power.
Pre-existing issues noticed in the surrounding code (not introduced by this patch):
- testcases/commands/nm/nm01.sh — Uses intermediate API (. tst_test.sh / tst_run) rather than the current style with tst_loader.sh and doc/env blocks. Not a blocker for this fix.
---
Note:
Our agent completed the review of the patch.
The agent can sometimes produce false positives although often its findings are genuine. If you find issues with the review, please comment this email or ignore the suggestions.
Regards,
LTP AI Reviewer
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
--
Andrea Cervesato
SUSE QE Automation Engineer Linux
[email protected]
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp
Hi Wei, Andrea,
FYI merged, thank you!
Kind regards,
Petr
Signature · 1 line
Mailing list info: https://lists.linux.it/listinfo/ltp