[
https://issues.apache.org/jira/browse/STDCXX-812?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12584373#action_12584373
]
Martin Sebor commented on STDCXX-812:
-------------------------------------
According to the test output (below), it spins 200,000 times by default:
{noformat}
# INFO (S1) (3 lines):
# TEXT: testing std::ctype<charT> with 1 thread, 200000 iterations each, in 16
locales { "C" "POSIX" "C.iso88591" "C.iso885915" "C.utf8" "univ.utf8"
"ar_DZ.arabic8" "ar_DZ.utf8" "ar_SA.arabic8" "ar_SA.iso88596" "ar_SA.utf8"
"bg_BG.iso88595" "bg_BG.utf8" "zh_CN.gb18030" "zh_CN.hp15CN" "zh_CN.utf8" }
# CLAUSE: lib.locale.ctype
{noformat}
In a narrow build (11s) the test takes just over a minute of real time to run
through all the 200,000 iterations:
{noformat}
tests$ time ./22.locale.ctype.mt >/dev/null
real 1m14.900s
user 0m57.390s
sys 0m17.040s
{noformat}
In a wide build (8D) though, it takes over a minute of real time just to run
20,000 iterations (I didn't have the patience to sit through the 200,000) so
ten times as many would take at least 10 minutes. The current timeout is set at
5 minutes (I think) so that explains the SIGHUP. The question is: why does the
test take 10 times longer to run in wide builds?
{noformat}
$ time ./22.locale.ctype.mt --nloops=20000 >/dev/null
real 1m11.879s
user 0m48.780s
sys 0m22.860s
{noformat}
> [HP aCC 6] timeouts in thread safety tests in wide (LP64) builds
> ----------------------------------------------------------------
>
> Key: STDCXX-812
> URL: https://issues.apache.org/jira/browse/STDCXX-812
> Project: C++ Standard Library
> Issue Type: Bug
> Components: Tests
> Affects Versions: 4.2.0
> Environment: HP aCC 6/HP-UX/IPF
> Reporter: Martin Sebor
> Assignee: Martin Sebor
> Fix For: 4.2.1
>
> Attachments: locale-tests.html
>
> Original Estimate: 4h
> Remaining Estimate: 4h
>
> A good number (but not all) of the locale thread safety tests are being
> killed with SIGHUP in nightly runs due to timeouts. That's odd since the
> thread safety tests typically sometimes time out in thread-safe builds but
> rarely (never?) in non-reentrant ones. This doesn't look like a regression
> since some of the same tests were timing out in 4.2.0 builds back in
> September 2007. We should at least try to figure out what's going on for
> 4.2.1 even if we can't take the time to fix it. Here's the relevant portion
> of the
> [hpux-11.31-ia64-acc-6.16-8S-642653-log.gz.txt|http://people.apache.org/~sebor/stdcxx/results/hpux-11.31-ia64-acc-6.16-8S-642653-log.gz.txt]
> run log with the head of trunk:
> {noformat}
> 22.locale.codecvt.length 0 0 408 0 100%
> 0.180 0.250 0.640
> 22.locale.codecvt.mt ABRT 0
> 3.660 3.210 7.800
> 22.locale.codecvt.out 0 0 1494 0 100%
> 0.120 0.160 0.390
> 22.locale.codecvt.stdcxx-435 NOUT 0
> 0.000 0.010 0.040
> 22.locale.cons.mt 0 0 0 0 100%
> 19.750 13.990 33.950
> 22.locale.cons.stdcxx-485 0 0 128 0 100%
> 0.070 0.090 0.250
> 22.locale.ctype.is 0 0 243880 0 100%
> 12.310 0.630 13.090
> 22.locale.ctype.mt HUP 0
> 181.550 117.320 300.020
> 22.locale.ctype.narrow 0 0 94062 0 100%
> 8.260 0.310 8.870
> 22.locale.ctype 0 0 876000 0 100%
> 26.980 1.360 28.550
> 22.locale.ctype.scan 0 0 164724 0 100%
> 1.110 0.430 1.630
> 22.locale.ctype.tolower 0 0 186668 0 100%
> 2.690 0.440 3.330
> 22.locale.ctype.toupper 0 0 186668 0 100%
> 2.680 0.430 3.220
> 22.locale.globals.mt 0 1 0 0 100%
> 51.020 29.220 80.670
> 22.locale.messages.mt 0 0 0 0 100%
> 2.570 4.970 7.770
> 22.locale.messages 0 0 366 0 100%
> 0.250 0.480 0.890
> 22.locale.messages.stdcxx-542 0 0 0 0 100%
> 0.010 0.010 0.100
> 22.locale.money.get.mt HUP 0
> 166.890 131.930 300.010
> 22.locale.money.get 0 0 3880 16 99%
> 0.030 0.020 0.140
> 22.locale.money.put.mt HUP 0
> 169.820 128.840 300.010
> 22.locale.money.put 0 0 1328 0 100%
> 0.010 0.010 0.090
> 22.locale.moneypunct.mt 0 0 0 0 100%
> 0.170 0.200 0.640
> 22.locale.moneypunct 0 0 12088 12 99%
> 0.370 0.370 0.820
> 22.locale.num.get.mt ABRT 0
> 0.110 0.150 0.760
> 22.locale.num.get 0 0 65412 96 99%
> 0.660 0.320 1.120
> 22.locale.num.put.mt HUP 0
> 169.090 129.710 300.010
> 22.locale.num.put 0 2 1671 0 100%
> 0.100 0.120 0.520
> 22.locale.num.put.stdcxx-2 NOUT 0
> 0.000 0.010 0.050
> 22.locale.numpunct.mt HUP 0
> 161.200 137.640 300.010
> 22.locale.statics.mt 0 0 0 0 100%
> 20.060 13.650 33.970
> 22.locale.stdcxx-554 NOUT 0
> 0.010 0.000 0.040
> 22.locale.time.get.mt 0 0 0 0 100%
> 144.670 93.470 239.160
> 22.locale.time.get 0 16 2092 46 97%
> 0.160 0.270 0.880
> 22.locale.time.put.mt HUP 0
> 161.430 137.530 300.020
> 22.locale.time.put 0 147 1854 13 99%
> 0.120 0.180 0.650
> {noformat}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
