ruby-core

Patches item #5111, was opened at 2006-07-18 14:36
You can respond by visiting: 
http://rubyforge.org/tracker/?func=detail&atid=1700&aid=5111&group_id=426

Category: Ruby1.8
>Group: v1.8.x
Status: Open
Resolution: None
Priority: 3
Submitted By: Jonathan Paisley (jzp)
Assigned to: Nobody (None)
Summary: Instability around popen due to missing rb_thread_atfork

Initial Comment:
A Ruby program with multiple threads making calls to IO.popen (or indirectly via backticks) can get some nasty race conditions if existing threads get a chance to run in the child.

This can happen, for example, if the rb_warn("Insecure world writable dir...") from file.c:path_check_0() gets invoked in the child.

Solution is to add a call to 'rb_thread_atfork()' in the child after fork() in order to kill off those other threads.

On a related theme, it seems like the 'rb_proc_exec' in process.c:pipe_open() should be protected with rb_protect or similar, to prevent any exceptions raised during rb_proc_exec() from being propagated to the soon-to-be-dead main thread in the child process. I'm not familiar enough with use of rb_protect to provide a proper patch for that.


Patch attached and included inline below:


diff -ur orig/ruby/io.c ruby/io.c
--- orig/ruby/io.c	2006-06-26 08:23:47.000000000 +0100
+++ ruby/io.c	2006-07-18 22:27:33.000000000 +0100
@@ -3102,6 +3102,7 @@
   retry:
     switch ((pid = fork())) {
       case 0:			/* child */
+	rb_thread_atfork();
 	if (modef & FMODE_READABLE) {
 	    close(pr[0]);
 	    if (pr[1] != 1) {


----------------------------------------------------------------------

Comment By: Jonathan Paisley (jzp)
Date: 2007-03-18 09:43

Message:
This problem still occurs in ruby-1.8.6. What can be done to get the simple 
patch to io.c into 1.8.x?

There was some discussion around the original submission with patches by 
Nobu Nakada:

   http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/8262


----------------------------------------------------------------------

You can respond by visiting: 
http://rubyforge.org/tracker/?func=detail&atid=1700&aid=5111&group_id=426

Thread

Prev Next

In This Thread

Prev Next

[#11066] question about frame_unique — Paul Brannan <pbrannan@...>

[#11069] High-precision Rationals cannot be converted to Floats — <noreply@...>

[#11070] Re: Welcome to our (ruby-core ML) You are added automatically — "Bunno Lim" <bunno_lim@...>

[#11071] autorunner -n option does not work in trunk — <noreply@...>

[#11073] segfault printing instruction sequence for iterator — <noreply@...>

[#11074] sub! causing unexpected side effect — <noreply@...>

[#11075] Ruby fails to parse valid wsdl — <noreply@...>

[#11076] Windows DEP causes crash — <noreply@...>

[#11077] Submitting an RDoc patch — gga <ggarra@...>

[#11078] sub! causing unexpected side effect — <noreply@...>

[#11079] ruby packer prototype: core dumps ( design review request ) — "Adam Bozanich" <adam.boz@...>

[#11081] sub! causing unexpected side effect — <noreply@...>

[#11082] Understanding code: Kernel#require and blocks. — Hugh Sasse <hgs@...>

[#11088] Net::HTTP encodes parameters improperly. — <noreply@...>

[#11090] parsedate fails — <noreply@...>

[#11092] return from block inside a lambda does not work as expected — <noreply@...>

[#11094] RDoc for Kernel#select same as IO.select — <noreply@...>

[#11095] RDoc for Kernel#select same as IO.select — <noreply@...>

[#11097] TestReadline fails depending on the terminal in which the test is run. — <noreply@...>

[#11098] Short read with readline, popen and timeout - first byte of the line disappears — <noreply@...>

[#11099] Short read with readline, popen and timeout - first byte of the line disappears — <noreply@...>

[#11100] Short read with readline, popen and timeout - first byte of the line disappears — <noreply@...>

[#11102] Net::HTTP encodes parameters improperly. — <noreply@...>

[#11103] irb attempts to escape '\c' inside single quotes — <noreply@...>

[#11104] overriding autoloads with classes and modules — Thomas Enebo <Thomas.Enebo@...>

[#11106] Failure parsing dates returned by WIN32OLE — <noreply@...>

[#11107] unable to build ruby HEAD without previously installed ruby — <noreply@...>

[#11108] pattern for implementation-private constants? — David Flanagan <david@...>

[#11117] BUGS in metaclasses inheritance — <noreply@...>

[#11118] BUGS in metaclasses inheritance — <noreply@...>

[#11121] Weird rounding behavior.. — <noreply@...>

[#11124] IRHG - Found interesting Japanese Document exploring various GC's — Charles Thornton <ceo@...>

[#11125] two fixes for sync.rb: sync_try_lock, Err.Fail — "Dmitry Borodaenko" <angdraug@...>

[#11127] Bugs that can be closed — "Jano Svitok" <jan.svitok@...>

[#11128] Weird rounding behavior.. — <noreply@...>

[#11129] sub! causing unexpected side effect — <noreply@...>

[#11130] :!~ not a symbol — <noreply@...>

[#11131] Documentation for String.scan has a bad example. — <noreply@...>

[#11132] rdoc/usage uses an uninitalized constant — <noreply@...>

[#11133] bug in newton.rb — <noreply@...>

[#11134] each_with_index (Enumerable) does not forward arguments to each — <noreply@...>

[#11135] Ruby Gem installer not playing nice on Ubuntu — <noreply@...>

[#11136] Ruby one-click installer wiped out system PATH environment variable. — <noreply@...>

[#11137] Ruby one-click installer wiped out system PATH environment variable. — <noreply@...>

[#11138] Marshal.dump shouldn't complain about singletons if the _dump method is defined — <noreply@...>

[#11139] exit from within block returns wrong value. — <noreply@...>

[#11140] :!~ not a symbol — <noreply@...>

[#11141] Weird rounding behavior.. — <noreply@...>

[#11143] Logger::Application accessor for logger — Brian Candler <B.Candler@...>

[#11144] Rational comparison to 0 fails when denominator is != 1 — <noreply@...>

[#11145] Rational comparison to 0 fails when denominator is != 1 — <noreply@...>

[#11147] RDoc patch to better support documenting multiple classes/SWIG — <noreply@...>

[#11150] "password" is case sensitive in lib/telnet.rb — <noreply@...>

[#11153] segfault printing instruction sequence for iterator — <noreply@...>

[#11154] Rational comparison to 0 fails when denominator is != 1 — <noreply@...>

[#11155] Class#=== bug — <noreply@...>

[#11156] Class#=== bug — <noreply@...>

[#11157] Class#=== bug — <noreply@...>

[#11158] Class#=== bug — <noreply@...>

[#11159] Rational comparison to 0 fails when denominator is != 1 — <noreply@...>

[#11165] Rational comparison to 0 fails when denominator is != 1 — <noreply@...>

[#11166] rdoc patches — Joel VanderWerf <vjoel@...>

[#11167] Date parsing bug with patch — <noreply@...>

[#11169] Allow back reference with nest level in Oniguruma for Ruby again — =?ISO-8859-15?Q?Wolfgang_N=E1dasi-Donner?= <wonado@...>

[#11171] Loss of precision during iteration. — <noreply@...>

[#11172] "read_nonblock" Problem (Windows and Ruby 1.8) — =?ISO-8859-15?Q?Wolfgang_N=E1dasi-Donner?= <wonado@...>

[#11174] Loss of precision during iteration. — <noreply@...>

[#11175] Loss of precision during iteration. — <noreply@...>

[#11176] FileUtils.rm_rf misfeature? — johan556@...

[#11177] Loss of precision during iteration. — <noreply@...>

[#11187] Garbage collection of Symbols in future release? — Sven Suska <sven716rc@...>

[#11193] IRHG -- Downloadable Files — Charles Thornton <ceo@...>

[#11204] URI Methods Raise Unclear Exceptions When Passed URI Objects — <noreply@...>

[#11206] ri HTML output bug with constants — Chris Thomas <chris@...>

[#11208] How to handle Change Proposals that affect mainly Oniguruma (for Ruby)? — =?ISO-8859-15?Q?Wolfgang_N=E1dasi-Donner?= <wonado@...>

[#11209] eval("0.0025 + 9 * 0.0025") === 0.025 => false — <noreply@...>

[#11210] Pathname ascend and descend inclusive parameter — TRANS <transfire@...>

[#11211] eval("0.0025 + 9 * 0.0025") === 0.025 => false — <noreply@...>

[#11213] eval(&quot;0.0025 + 9 * 0.0025&quot;) === 0.025 =&gt; false — <noreply@...>

[#11214] eval(&quot;0.0025 + 9 * 0.0025&quot;) === 0.025 =&gt; false — <noreply@...>

[#11213] eval("0.0025 + 9 * 0.0025") === 0.025 => false — <noreply@...>

[#11214] eval("0.0025 + 9 * 0.0025") === 0.025 => false — <noreply@...>

[#11216] eval("0.0025 + 9 * 0.0025") === 0.025 => false — <noreply@...>

[#11290] [TINY][PATCH] Array#last is not working for n > 3 — <noreply@...>

[#11291] In a c++ program, when I call "fclose()", or "ofstream::close()" in a file that includes ruby.h, the program crashes — <noreply@...>

[#11318] Problem with Benchmark::Tms#add! - expects &blk but takes no parameters — <noreply@...>