[#9722] Kernel#system broken inside Dir.chdir(&block) if system command doesn't have shell characters — <noreply@...>

Bugs item #7278, was opened at 2006-12-14 13:59

8 messages 2006/12/14

[#9749] System V IPC in standard library? — Steven Jenkins <steven.jenkins@...>

Back in August, I needed a semaphore to serialize access to an external

14 messages 2006/12/19

[#9753] CVS freeze — SASADA Koichi <ko1@...>

Hi,

20 messages 2006/12/20
[#9755] Re: [ruby-dev:30039] CVS freeze — SASADA Koichi <ko1@...> 2006/12/20

Hi,

[#9757] Re: [ruby-dev:30040] Re: CVS freeze — SASADA Koichi <ko1@...> 2006/12/20

Hi,

Re: Ruby performance improvements

From: khaines@...
Date: 2006-12-16 11:20:49 UTC
List: ruby-core #9734
On Tue, 5 Dec 2006, Michael Selig wrote:

> Anyway, I have recreated the patch (attached) and tested that it works!

I applied this patch for testing and discovered that the version that it 
sets, 1.8.4-OPT, messes up part of Ruby's build process where it breaks 
down the major and minor numbers of the version in order to use them to 
form directory names.

So I tweaked it to say that it was 1.8.4 and built using these patches.

I tested it with a fairly sophisticated web site running under IOWA, and 
have only found one problem.

An attr_accessor call with a long argument list (25 symbols) only received 
the first 17 (if I counted right).  The remaining 8 or so were ignored.
I broke that call into 2 calls, each with 1/2 the symbols, and retested.

It seems to work just fine, with an approximately 15% performance increase 
over a standard 1.8.4 installation.  I ran about 60000 requests through 
the application that I was testing with, and I didn't see any unusual 
memory behavior or inconsistent results.

15% isn't earthshaking, but it's certainly not bad if nothing in a fairly 
complex set of code breaks (other than the one aforementioned item) to 
get it.


Kirk Haines

In This Thread