win32utils devel - Oct 2007 - Simulating RUBY_CRITICAL

Hi all,

I''ve added some critical section methods to the Windows::Synchronize 
module in the windows-pr library (in CVS). Looking at the RUBY_CRITICAL 
macro from rubysig.h, it basically looks like this (in pure Ruby):

def RUBY_CRITICAL(&block)
    critical_section = [0].pack(''L'')
    InitializeCriticalSection(critical_section)
    EnterCriticalSection(critical_section)
    block.call
    LeaveCriticalSection(critical_section)
end

But, shouldn''t we ensure that we leave the critical section once 
entered, like so?

# This is what''s currently in CVS
def RUBY_CRITICAL(&block)
    critical_section = [0].pack(''L'')
    begin
       InitializeCriticalSection(critical_section)
       EnterCriticalSection(critical_section)
       block.call
    ensure
       LeaveCriticalSection(critical_section)
    end
end

Or, am I worrying over nothing?

Regards,

Dan

Hi Dan,

I certainly like the idea of ensuring that we leave the critical
section once entered.  However, there are some possible complications.
 For example, MSDN says:

"Windows 2000/NT:  In low memory situations, EnterCriticalSection can
raise an exception. Do not attempt to handle this exception; instead,
either terminate the process or allow the exception to pass to the
unhandled exception filter. To avoid an exception due to low memory,
call the InitializeCriticalSectionAndSpinCount function to preallocate
the event used by EnterCriticalSection instead of calling the
InitializeCriticalSection function, which forces EnterCriticalSection
to allocate the event. Preallocating the event is not necessary on
Windows XP or later because EnterCriticalSection is guaranteed not to
fail due to lack of resources."
http://msdn2.microsoft.com/en-us/library/ms682608.aspx

Then again, I don''t think it''s too likely that we''d
be low enough on
memory for EnterCriticalSection to fail.

Best regards,

Wayne

On 10/20/07, Daniel Berger <djberg96 at gmail.com>
wrote:
> Hi all,
>
> I''ve added some critical section methods to the
Windows::Synchronize
> module in the windows-pr library (in CVS). Looking at the RUBY_CRITICAL
> macro from rubysig.h, it basically looks like this (in pure Ruby):
>
> def RUBY_CRITICAL(&block)
>     critical_section = [0].pack(''L'')
>     InitializeCriticalSection(critical_section)
>     EnterCriticalSection(critical_section)
>     block.call
>     LeaveCriticalSection(critical_section)
> end
>
> But, shouldn''t we ensure that we leave the critical section once
> entered, like so?
>
> # This is what''s currently in CVS
> def RUBY_CRITICAL(&block)
>     critical_section = [0].pack(''L'')
>     begin
>        InitializeCriticalSection(critical_section)
>        EnterCriticalSection(critical_section)
>        block.call
>     ensure
>        LeaveCriticalSection(critical_section)
>     end
> end
>
> Or, am I worrying over nothing?
>
> Regards,
>
> Dan
> _______________________________________________
> win32utils-devel mailing list
> win32utils-devel at rubyforge.org
> http://rubyforge.org/mailman/listinfo/win32utils-devel
>

Wayne Vucenic wrote:
> Hi Dan,
> 
> I certainly like the idea of ensuring that we leave the critical
> section once entered.  However, there are some possible complications.
>  For example, MSDN says:
> 
> "Windows 2000/NT:  In low memory situations, EnterCriticalSection can
> raise an exception. Do not attempt to handle this exception; instead,
> either terminate the process or allow the exception to pass to the
> unhandled exception filter. To avoid an exception due to low memory,
> call the InitializeCriticalSectionAndSpinCount function to preallocate
> the event used by EnterCriticalSection instead of calling the
> InitializeCriticalSection function, which forces EnterCriticalSection
> to allocate the event. Preallocating the event is not necessary on
> Windows XP or later because EnterCriticalSection is guaranteed not to
> fail due to lack of resources."
> http://msdn2.microsoft.com/en-us/library/ms682608.aspx
> 
> Then again, I don''t think it''s too likely that
we''d be low enough on
> memory for EnterCriticalSection to fail.
<snip>

Hi Wayne,

Yeah, I saw that. I guess I''m not especially worried about checking for
low memory for two reasons. First, it seems we don''t have to worry
about
it with XP and later, which covers the majority of Windows users at this 
point I think. Second, I''m pretty sure the interpreter would die anyway
if there were a memory allocation issue.

But, now that I look at this code, I''m somewhat confused as to why they
felt compelled to wrap methods like Dir.mkdir in a RUBY_CRITICAL block 
in the first place. Are they trying to make it thread safe? The Ruby 
interpreter isn''t thread safe to begin with. Are they trying to make 
create + chmod a single transaction? Seems overkill, and besides, the 
code in dir.c takes no such precaution for Unix.

I''d really like to see the use cases for why this was used.

Regards,

Dan