Bug #2048
closed
Thread#raise: Handling of Current Exception
Added by runpaint (Run Paint Run Run) almost 15 years ago.
Updated over 13 years ago.
Description
=begin
$ cat /tmp/thread.rb
t = Thread.new do
begin
1/0
rescue ZeroDivisionError
sleep 3
end
end
Thread.pass while t.status and t.status != "sleep"
t.raise
t.value
t.kill
$ ruby -v /tmp/thread.rb
ruby 1.9.2dev (2009-09-03 trunk 24741) [i686-linux]
/tmp/thread.rb:5:in sleep': unhandled exception from /tmp/thread.rb:5:in
rescue in block in '
from /tmp/thread.rb:2:in `block in '
$ ruby8 -v /tmp/thread.rb
ruby 1.8.8dev (2009-09-04) [i686-linux]
/tmp/thread.rb:3:in /': divided by 0 (ZeroDivisionError) from /tmp/thread.rb:11:in
value'
from /tmp/thread.rb:11
On 1.8 Thread#raise re-raises the current exception; on 1.9 it doesn't.
=end
- Status changed from Open to Assigned
- Assignee set to ko1 (Koichi Sasada)
- Priority changed from Normal to 5
- Target version set to 1.9.2
- Status changed from Assigned to Closed
- % Done changed from 0 to 100
=begin
Applied in changeset r25278.
=end
- Status changed from Closed to Open
=begin
I think r25278 is wrong.
The point should be that rescue clause restores its originating exception in 1.8 but it doesn't in 1.9.
Still I'm not sure if this is intentional change.
=end
- Status changed from Open to Closed
=begin
Sorry, it comes feeling right.
And seems rdoc doesn't mention the no argument case.
=end
=begin
If you change interpreter or VM of Ruby, please get an approval by matz in ruby-core/Redmine (or ruby-dev).
Sorry for this was an implicit rule,
but it is important to let core developers know such critical changes.
Anyway, can you fold your commit logs by about 80 columns?
=end
=begin
Hi,
In message "Re: [ruby-core:26056] Re: [Bug #2048] Thread#raise: Handling of Current Exception"
on Mon, 12 Oct 2009 13:54:21 +0900, Marc-Andre Lafortune ruby-core-mailing-list@marc-andre.ca writes:
|> The 1.8 behavior is an accidental feature, so that it should not be
|> used nor documented. I should fix 1.8 rather.
|
|Oh.
|So should Thread#raise require at least one argument? Or else
|Thread#raise takes the exception being handled by the current thread
|and raises it in the receiving thread?
I'd rather leave it implementation dependent, but my intention is:
when Thread#raise is called without any argument, it would raise
RuntimeError without any message (as 1.9 does).
matz.
=end
Also available in: Atom
PDF
Like0
Like0Like0Like0Like0Like0Like0Like0