| Summary: | gtod.exp fails on ppc64/i386 | ||
|---|---|---|---|
| Product: | systemtap | Reporter: | Srinivasa DS <srinivasa> |
| Component: | testsuite | Assignee: | Unassigned <systemtap> |
| Status: | RESOLVED WORKSFORME | ||
| Severity: | normal | CC: | mhiramat, wcohen |
| Priority: | P2 | ||
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Host: | Target: | ||
| Build: | Last reconfirmed: | ||
|
Description
Srinivasa DS
2007-10-03 06:27:09 UTC
I noticed a similar failure on an i686 machine running a stock kernel,
2.6.18-53.1.4EL with the current stap_snap_testing (20080112). The gtod.exp
test has the following matching:
-re {^[0-9]+ \:([0-9]+) appl\r\n[0-9]+ \:\1 kern\r\n[0-9]+ \:\1 prog\r\n} {
incr ok; exp_continue }
the \1 before "kern" and "prog" expect need something like the following to match:
1200411986605858 :00 appl
1200411986605862 :00 kern
1200411986605873 :00 prog
However, there are places in the output where the number entries are not grouped
together (notice the 78 before the last 77 and the 79 between the 78's):
1200411986606250 :77 kern
1200411986606252 :77 appl
1200411986606255 :78 kern
1200411986606256 :77 prog
1200411986606257 :78 appl
1200411986606260 :79 kern
1200411986606261 :78 prog
1200411986606262 :79 appl
1200411986606265 :79 prog
It appears that the get_time_of_day is not being monotonic. The particular i686
machine was a laptop, which could have changes in the clock frequency. I wonder
if the hypervisor on the ppc64 could be skewing the results.
Seeing something similar on 2.6.27.21-170.2.56.fc10.i686 (SMP) 1240479726271339 :10 appl 1240479726271341 :10 kern 1240479726271347 :10 prog 1240479726281401 :11 appl 1240479726281403 :11 kern 1240479726281409 :11 prog 1240479726291442 :12 kern 1240479726291494 :12 appl 1240479726291501 :12 prog 1240479726301442 :13 kern 1240479726301586 :13 appl 1240479726301594 :13 prog 1240479726311442 :14 kern 1240479726311678 :14 appl 1240479726311686 :14 prog Notice how the sequence is not always appl-kern-prog, sometimes it is kern-appl-prog. Seeing the gtod.exp test passing on both ppc64le fc28. [wcohen@rh-power-vm07 systemtap]$ uname -r 5.0.16-100.fc28.ppc64le [wcohen@rh-power-vm07 systemtap]$ grep gtod testsuite/systemtap.sum Running /home/wcohen/systemtap_write/systemtap/testsuite/systemtap.base/gtod.exp ... PASS: gtod (short range) (100) PASS: gtod (10ms interval) (100) PASS: gtod (100ms interval) (100) And and i386 f30: [wcohen@weci686 testsuite]$ uname -r 5.0.17-300.fc30.i686 [wcohen@weci686 testsuite]$ grep gtod systemtap.sum Running /usr/share/systemtap/testsuite/systemtap.base/gtod.exp ... PASS: gtod (short range) (100) PASS: gtod (10ms interval) (100) PASS: gtod (100ms interval) (100) |