e$B$-$7$b$H$G$9e(B
http://redmine.ruby-lang.org/issues/show/270 e$B$K$b$"$ke(B
e$B!Ve(BFreeBSDe$B$de(BNetBSDe$B$O!$0lEYe(B pthread
e$B$r:n$k$H!$8e;OKv$r$7$Fe(B
e$B$be(B fork e$B8e!$e(Bpthread
e$B$r:n$m$&$H$9$k$H;I$5$k!W$K4X$9$k7o$J$N$G$9$,!"e(B
e$B8D?ME*$K;H$C$F$$$k%D!<%k$,$3$l$N$?$a$Ke(B Ruby1.9
e$B$K0\9T$G$-$J$$$G$$$^$9!#e(B
e$B$3$l$Oe(B OS e$BB&$N%P%0$J$s$G$7$g$&$+!"$=$l$H$be(B POSIX
e$BE*$K$7$g$&$,$J$$e(B
e$B;EMM$J$s$G$7$g$&$+!#e(B
FreeBSD e$B$Ne(B send-pr e$B$re(B thread
e$B$G8!:w$7$F$_$?46$8$G$O!“Js9p$5$l$Fe(B
e$B$J$5$=$&$J$N$G!”%P%0$H$$$&$3$H$J$iJs9p$7$?$$$N$G$9$,!#e(B
At Tue, 31 Mar 2009 18:48:46 +0900,
KISHIMOTO, Makoto wrote:
http://redmine.ruby-lang.org/issues/show/270 ã«ã‚‚ã‚ã‚‹
「FreeBSDã‚„NetBSDã¯ï¼Œä¸€åº¦ pthread を作るã¨ï¼Œå¾Œå§‹æœ«ã‚’ã—ã¦
ã‚‚ fork 後，pthread を作ã‚ã†ã¨ã™ã‚‹ã¨åˆºã•ã‚‹ã€ã«é–¢ã™ã‚‹ä»¶ãªã®ã§ã™ãŒã€
個人的ã«ä½¿ã£ã¦ã„るツールãŒã“ã‚Œã®ãŸã‚ã« Ruby1.9 ã«ç§»è¡Œã§ããªã„ã§ã„ã¾ã™ã€‚
ã“れ㯠OS å´ã®ãƒã‚°ãªã‚“ã§ã—ょã†ã‹ã€ãã‚Œã¨ã‚‚ POSIX çš„ã«ã—ょã†ãŒãªã„
仕様ãªã‚“ã§ã—ょã†ã‹ã€‚
FreeBSD ã® send-pr ã‚’ thread ã§æ¤œç´¢ã—ã¦ã¿ãŸæ„Ÿã˜ã§ã¯ã€å ±å‘Šã•ã‚Œã¦
ãªã•ãã†ãªã®ã§ã€ãƒã‚°ã¨ã„ã†ã“ã¨ãªã‚‰å ±å‘Šã—ãŸã„ã®ã§ã™ãŒã€‚
ã€€å ±å‘Šã•ã‚Œèªè˜ã‚‚ã•ã‚Œã¦ã„ã¾ã™ãŒã€æ®‹å¿µãªãŒã‚‰ä»Šã®ã¨ã“ã‚ä¿®æ£äºˆå®šã¯
ã‚ã‚Šã¾ã›ã‚“。POSIX/Pthread ã®ä»•æ§˜çš„ã«ã‚‚動作ä¿è¨¼å¤–ãªã®ã§ã€å¿…æ»ã«
é ‘å¼µã£ã¦ã¾ã§ä½•ã¨ã‹ã—よã†ã¨ã„ã†å‹•ãã«ã¯ãªã£ã¦ã„ãªã„よã†ã§ã™ã€‚
http://www.freebsd.org/cgi/query-pr.cgi?pr=threads/101323
　JDK/CPythonç‰ã€ä»–ã®è¨€èªžå‡¦ç†ç³»ã¨ã¯ä¼¼ãŸå•é¡Œã‚’共有ã—ã¦ã„ã‚‹ã®ã§ã€
ãã¡ã‚‰ã«èª¿æŸ»ã‚’広ã’ã‚‹ã¨ã„ã„ã‹ã‚‚ã—ã‚Œã¾ã›ã‚“。
　ã¾ãŸã€å…·ä½“çš„ãªã‚„ã‚ŠãŸã„ã“ã¨ãŒç¤ºã•ã‚Œã‚Œã°ä»£æ›¿æ‰‹æ®µã‚’æ示ã§ãã‚‹ã‹ã‚‚
ã—ã‚Œã¾ã›ã‚“。fork+signal ã‚’ spawn+何らã‹ã®IPC ã«ç½®ãæ›ãˆã‚‹ã¨ã‹ã€
daemon ã‚’ daemontools ã‚„ (Macãªã‚‰) launchd ãªã©ã®å¤–部ツールã«
ä»»ã›ã‚‹ã¨ã‹ã€‚(ã¡ã‚‡ã£ã¨æ‚”ã—ã„ã§ã™ãŒ)
In message [email protected]
on Tue, 31 Mar 2009 18:48:46 +0900,
“KISHIMOTO, Makoto” [email protected] wrote:
http://redmine.ruby-lang.org/issues/show/270 e$B$K$b$“$ke(B
e$B!Ve(BFreeBSDe$B$de(BNetBSDe$B$O!$0lEYe(B pthread e$B$r:n$k$H!$8e;OKv$r$7$Fe(B
e$B$be(B fork e$B8e!$e(Bpthread e$B$r:n$m$&$H$9$k$H;I$5$k!W$K4X$9$k7o$J$N$G$9$,!“e(B
…
e$B$3$l$Oe(B OS e$BB&$N%P%0$J$s$G$7$g$&$+!”$=$l$H$be(B POSIX e$BE*$K$7$g$&$,$J$$e(B
e$B;EMM$J$s$G$7$g$&$+!#e(B
FreeBSD e$B$Ne(B send-pr e$B$re(B thread e$B$G8!:w$7$F$_$?46$8$G$O!“Js9p$5$l$Fe(B
e$B$J$5$=$&$J$N$G!”%P%0$H$$$&$3$H$J$iJs9p$7$?$$$N$G$9$,!#e(B
e$B40A4$KM}2r$7$F$$$k$o$1$G$O$”$j$^$;$s$,!"e(BNetBSDe$B$Ne(Bpthread_atfork(3)e$B$K!"e(B
CAVEATS
After calling fork(2) from a multithreaded process, it is only safe
to
call async-signal-safe functions until calling one of the exec(3)
func-
tions. The pthread_*() functions are not async-signal-safe, so it
is not
safe to use such functions in the child handler.
e$B$J$s$F=q$+$l$F$$$k$“$?$j$b4X78$”$j$=$&$K;W$($^$9!#e(B
In article [email protected],
Urabe S. [email protected] writes:
e$B$A$g$C$H$3$N!Ve(BPthreade$B$N;EMME*$K$bF0:nJ]>c30!W$C$FItJ,$N:,5r$,$I$&$,$s$P$C$F$b<+e(B
e$BNO$GH/8+$G$-$J$$$N$G!"$h$m$7$1$l$P2?$+%]%$%s%?$r<($7$F$$$?$@$1$^$;$s$+e(B?
fork e$B$N9`$K$"$ke(B
Consequently, to avoid errors, the child process may only
execute async-signal-safe operations until such time as
one of the exec functions is called.
fork
e$B$H$$$&$H$3$m$J$s$8$c$J$$$+$H;W$$$^$9!#e(B
async-signal-safe function e$B$N%j%9%H$Oe(B
http://www.opengroup.org/onlinepubs/009695399/functions/xsh_chap02_04.html
e$B$K:$C$F$$$F!"e(Bpthread_create e$B$OF~$C$F$$$^$;$s!#e(B
e$B$J$N$G!"5,3J$H$7$F$O!"e(Bfork e$B$G$G$-$?;R%W%m%;%9$G$Oe(B exec
e$B$9$ke(B
e$B$^$Ge(B pthread_create e$B$r;H$C$F$O$$$1$^$;$s!#e(B
e$BKNIt$G$9!#e(B
Akinori MUSHA e$B$5$s$O=q$-$^$7$?e(B:
e$B$3$l$Oe(B OS e$BB&$N%P%0$J$s$G$7$g$&$+!"$=$l$H$be(B POSIX e$BE*$K$7$g$&$,$J$$e(B
e$B;EMM$J$s$G$7$g$&$+!#e(B
FreeBSD e$B$Ne(B send-pr e$B$re(B thread e$B$G8!:w$7$F$_$?46$8$G$O!“Js9p$5$l$Fe(B
e$B$J$5$=$&$J$N$G!”%P%0$H$$$&$3$H$J$iJs9p$7$?$$$N$G$9$,!#e(B
e$B!!Js9p$5$lG’<1$b$5$l$F$$$^$9$,!“;DG0$J$,$i:#$N$H$3$m=$@5M=Dj$Oe(B
e$B$”$j$^$;$s!#e(BPOSIX/Pthread e$B$N;EMME*$K$bF0:nJ]>Z30$J$N$G!"I,;`$Ke(B
e$B4hD%$C$F$^$G2?$H$+$7$h$&$H$$$&F0$-$K$O$J$C$F$$$J$$$h$&$G$9!#e(B
e$B$A$g$C$H$3$N!Ve(BPthreade$B$N;EMME*$K$bF0:nJ]>c30!W$C$FItJ,$N:,5r$,$I$&$,$s$P$C$F$b<+e(B
e$BNO$GH/8+$G$-$J$$$N$G!“$h$m$7$1$l$P2?$+%]%$%s%?$r<($7$F$$$?$@$1$^$;$s$+e(B?
pthread_atfork()e$B$H$+$$$&4X?t$,$”$k$h$&$Ge(B
http://www.opengroup.org/onlinepubs/009695399/functions/pthread_atfork.html
e$B$5$9$,$KF0:nJ]>c30$C$F$3$H$O$J$$$s$8$c$J$$$+$H;W$&$s$G$9$1$I$b!#e(B
e$B$-$7$b$H$G$9e(B
e$B3’MM$"$j$,$H$&$4$6$$$^$9!#e(Bspawne$B$G$d$C$F$_$^$9!#e(B
e$B;R%W%m%;%9$G!“I8=F~NO$r?F$+$i0z$-7Q$$$@%Q%$%W$K$D$J$.$+$($F$+$ie(Bexece$B$7$F$ke(B e$B!J=PNO$rJL%W%m%0%i%$N%U%#%k%?$KDL$7$F$k!K$N$G!”%7%'%k$N%U%!%$%k%G%#%9%/%j%W%?$re(B
e$B;XDj$7$?%j%@%$%l%/%H$r;H$C$F$G$-$J$$$+$d$C$F$_$^$9!#B?J,$G$-$=$&!#e(B
forke$B$r;H$C$F$k$H;W$$9~$s$G$?=j$G$O;H$C$F$$$J$/$F!"e(B
e$BA4$/K:$l$F$?ItJ,$G;H$C$F$$$?e(Borz
PS. Ruby
spawne$B$G8!:w$7$?$i!":#7ne(B15e$BF|$N%;%%J!<$r$$D$1$^$7$?!#e(B
http://www.fsij.org/homepage/wiki/MonthlyMeeting2009Apr
Urabe S. e$B$5$s$O=q$-$^$7$?e(B:
one of the exec functions is called.
e$B$^$Ge(B pthread_create e$B$r;H$C$F$O$$$1$^$;$s!#e(B
e$B$J$k$[$I!#$"$j$,$H$&$4$6$$$^$9!#e(B
e$B$7$+$7$9$k$He(Bpthread_atforke$B$Ne(BRATIONALEe$B$K$"$k!Ve(Bforke$B$NA0$Ke(Bmutexe$B$r3NJ]$7$Fe(Bforke$B8e$Ke(B
e$B3+J|$9$k!W$C$F$d$D$O$=$b$=$bL5M}$J$s$G$9$M$(!#e(Bpthread_mutex_unlocke$B$be(Basync-
signal-safee$B$G$O$J$$$h$&$G$9$7!#e(B
Tanaka A. e$B$5$s$O=q$-$^$7$?e(B:
one of the exec functions is called.
e$B$^$Ge(B pthread_create e$B$r;H$C$F$O$$$1$^$;$s!#e(B
e$B$J$k$[$I!#$"$j$,$H$&$4$6$$$^$9!#e(B
In article [email protected],
“KISHIMOTO, Makoto” [email protected] writes:
e$B;R%W%m%;%9$G!“I8=F~NO$r?F$+$i0z$-7Q$$$@%Q%$%W$K$D$J$.$+$($F$+$ie(Bexece$B$7$F$ke(B e$B!J=PNO$rJL%W%m%0%i%$N%U%#%k%?$KDL$7$F$k!K$N$G!”%7%'%k$N%U%!%$%k%G%#%9%/%j%W%?$re(B
e$B;XDj$7$?%j%@%$%l%/%H$r;H$C$F$G$-$J$$$+$d$C$F$_$^$9!#B?J,$G$-$=$&!#e(B
forke$B$r;H$C$F$k$H;W$$9~$s$G$?=j$G$O;H$C$F$$$J$/$F!"e(B
e$BA4$/K:$l$F$?ItJ,$G;H$C$F$$$?e(Borz
e$B$$=$i$/%7%‘%k$OITMW$@$H;W$$$^$9$h!#e(B
(e$B%7%’%k<+BN$rF0$+$9$N$,L\E$G$J$$8B$j$Oe(B)
spawn e$B$N%^%K%e%"%ke(B (rdoc) e$B$rFI$s$G$_$F$/$@$5$$!#e(B
e$B$J$*!“e(B1.9.2 e$B$Ne(B open3
e$B$K$O!”%Q%$%W%i%$%s$J$I$“$j$,$A$J5/F0$Ne(B
e$B$7$+$?$O4JC1$K9T$($k%a%=%C%I$rIU$1$F$”$k$N$G!"Lr$KN)$D$+$b$7e(B
e$B$l$^$;$s!#e(B
PS. Ruby spawne$B$G8!:w$7$?$i!":#7ne(B15e$BF|$N%;%%J!<$r$$D$1$^$7$?!#e(B
http://www.fsij.org/homepage/wiki/MonthlyMeeting2009Apr
e$BOC<T$O;d$G$9!#6=L#$,$"$k$+$?$O$I$&$>!#e(B
In article [email protected],
“KISHIMOTO, Makoto” [email protected] writes:
Pythone$B$Ne(Bsubprocesse$B$_$?$$$J5!G=$,DI2C$5$l$F$?$s$G$9$M!#e(B
e$B$O$$!"e(BPython e$B$Ne(B subprocess e$B$O;29M$K$7$^$7$?!#e(B
e$B7k2L$H$7$F$G$-$?e(B API
e$B$,;w$F$$$k$+$I$&$+$O2x$7$$$H$3$m$G$9$,!"e(B
e$B$I$A$i$b%f!<%6$NL\E*$rD>@\E*$K;Y1g$9$k$h$&$K0U?^$7$F$$$k$H;We(B
e$B$$$^$9!#e(B
spawn e$B$N%^%K%e%"%ke(B (rdoc) e$B$rFI$s$G$_$F$/$@$5$$!#e(B
e$B$"$j$,$H$&$4$6$$$^$9!#%j%U%!%l%s%9%^%K%e%"%k$7$+FI$s$G$^$;$s$G$7$?!#e(B
Pythone$B$Ne(Bsubprocesse$B$_$?$$$J5!G=$,DI2C$5$l$F$?$s$G$9$M!#e(B
e$B$J$<$+%7%’%k7PM3$G$O$&$^$/$$$+$J$+$C$?$s$G$9$,!"e(Bspawne$B$N%*%W%7%g%s0z?t$re(B
e$B;H$C$F$&$^$/$$$-$^$7$?!#e(B
e$B$J$*!“e(B1.9.2 e$B$Ne(B open3 e$B$K$O!”%Q%$%W%i%$%s$J$I$"$j$,$A$J5/F0$Ne(B
e$B$7$+$?$O4JC1$K9T$($k%a%=%C%I$rIU$1$F$"$k$N$G!"Lr$KN)$D$+$b$7e(B
e$B$l$^$;$s!#e(B
waitpide$B$GBT$D$h$&$K$7$?$+$C$?$N$G!“e(Bpide$B$,<h$l$J$$e(Bpopene$B7O$O;H$($J$$$He(B
e$BH=CG$7$?<BAu;~$N5-21$,$”$k$N$G$9$,!"$3$A$i$b:#$Oe(Bpide$B$,<h$l$k$h$&$K$J$C$Fe(B
e$B$$$k$N$G$9$M!#e(Bopen3e$B$r;H$&$[$&$b<B83$7$F$_$^$9!#e(B
waitpide$B$GBT$D$h$&$K$7$?$+$C$?$N$G!“e(Bpide$B$,<h$l$J$$e(Bpopene$B7O$O;H$($J$$$He(B
e$BH=CG$7$?<BAu;~$N5-21$,$”$k$N$G$9$,!"$3$A$i$b:#$Oe(Bpide$B$,<h$l$k$h$&$K$J$C$Fe(B
e$B$$$k$N$G$9$M!#e(Bopen3e$B$r;H$&$[$&$b<B83$7$F$_$^$9!#e(B
Open3.pipeline_w e$B$,$&$^$/;H$($^$7$?!#e(B
waitpid e$B$Oe(B wait_thread e$B$K$b$C$F$+$l$F$7$^$&$N$G!"85$Oe(B
waitpid
e$B$@$C$?$H$3$m$Oe(B wait_thread.value e$B$GBT$D$h$&$K$7$^$7$?!#e(B