TheCodeShop Google Group

Re: [TheCodeShop] New Ruby 1.9.3-p429 with security fixes

beltranr...@gmail.com (Vdoble) — Tue, 21 May 2013 16:44:41 UT

Hi Jon,
Thanks for the detailed email. Yes, it is not a problem to compile the new
version with the patches on Linux but I trying to find a compiled version
for Windows. We will check if it is possible to move to 2.0.
Thanks again.

Re: [TheCodeShop] New Ruby 1.9.3-p429 with security fixes

jon.for...@gmail.com (Jon) — Sun, 19 May 2013 18:13:07 UT

What is your primary OS? Regardless of whether you're on Linux, Windows, or OS X, if you're not limited to 1.9.3 for some reason, I'd move to 2.0.0.

If you still need 1.9.3 it looks like Yura's made updates to his Falcon patches so you may be able to patch up a plain vanilla Ruby 1.9.3 and build from source. If Yura's still monitoring this ML, he may have more specific advice on how best to use his 1.9.3 backport patches.

Re: [TheCodeShop] New Ruby 1.9.3-p429 with security fixes

beltranr...@gmail.com (Vdoble) — Sun, 19 May 2013 16:58:32 UT

Thanks Jon,
As far as I understand, the Code Shop Ruby version ships some patches for
Ruby 1.9 version that improves the time startup significantly, for example
Falcon's patch [link] or Greg Price's
patch [link] Would be possible to get a new

Re: [TheCodeShop] New Ruby 1.9.3-p429 with security fixes

jon.for...@gmail.com (Jon) — Fri, 17 May 2013 13:09:43 UT

On Fri, 17 May 2013 02:13:22 -0700 (PDT)

Hi Vdoble,

No new experimental versions of TCS Ruby are planned because many of the improvements have already been included upstream
in the official Ruby 1.9 and 2.x branches. As such, I consider TCS' MRI project [link]

New Ruby 1.9.3-p429 with security fixes

beltranr...@gmail.com (Vdoble) — Fri, 17 May 2013 09:13:22 UT

A new Ruby 1.9 version was released recently to address a security issue
[link]
Do you plan to build a new version of Ruby? Thanks.

[ANN] TCS 1.9.3-p??? experimental build

jon.for...@gmail.com (Jon) — Sat, 25 Aug 2012 00:01:19 UT

TCS Ruby 1.9.3 => MRI 1.9.3 performance optimizations for Windows and *nix users.

This is primarily a maintenance release to get the latest MRI 1.9.3 backports. Yura's optimizations (require speedups, Hash, and backported 2.0.0 GC for *nix systems) also remain integrated.

As there have been no reported Fenix-related issues, and the ruby-core tests continue to pass on both Windows and *nix, I'm retiring the `fenix\ruby_1_9_3` (e15806e) and `fileload_ok/ruby_1_9_3` (eab4d19) patch branches originally used to enable optional Fenix support from an external module. These two patch branches have been subsumed by the `win-file/ruby_1_9_3` patch branch maintained by Hiroshi.

[alpha] uru the multi-platform ruby wrangler

jon.for...@gmail.com (Jon) — Tue, 07 May 2013 04:35:49 UT

I recently dusted off an old toy project I'd created while spelunking go in the pre-1.0 days. Amazingly, it's now building on go's 1.1 tip.

[link]

Uru's got the beginnings of documentation

[link]

and what's likely to blossom into a spectacularly large and tricky list of wacky system issues

Re: [TheCodeShop] Cry for help regarding with the WDM gem issue.

jarm...@gmail.com (Jarmo Pertman) — Mon, 21 Jan 2013 17:28:44 UT

You guys are just awesome :)
I can confirm that this change fixes the leak on my Win8 64bit machine too!
Thank you Hiroshi for detecting and fixing the problem. And Jon, thank you
for such a detailed write-up.
I've created a pull request [1] for wdm and mentioned you guys there too :)
[1] - [link]

Re: [TheCodeShop] Cry for help regarding with the WDM gem issue.

jon.for...@gmail.com (Jon) — Mon, 21 Jan 2013 05:32:39 UT

Just got back in, fired up email and see that Hiroshi-san delivered the awesome good-fu once again! :) :) :)

Jarmo: you likely already know the ruby memory mgmt code Hiroshi described, but if not, here's a snippet walkthrough for future reference. Hiroshi, please correct me if I've mislead.

Re: [TheCodeShop] Cry for help regarding with the WDM gem issue.

h.shiros...@gmail.com (Hiroshi Shirosaki) — Mon, 21 Jan 2013 02:04:56 UT

Hello,

[link]

When adding `GC.disable` to the test script, "[FATAL] failed to
allocate memory" doesn't occur on my test.
When adding `GC.stress = true`, "[FATAL] failed to allocate memory"
occurs immediately.

So this issue would be related to gc and rb_thread_blocking_region

Re: [TheCodeShop] Cry for help regarding with the WDM gem issue.

jon.for...@gmail.com (Jon) — Sun, 20 Jan 2013 15:37:44 UT

I'll try to find time to fire up GDB late today/tommorrow when I can look at this again, but it appears your sample code is causing ruby to fail in one of the following (ruby_1_9_3 branch) locations:

gc.c:572
gc.c:584
gc.c:739
thread_pthread.c
vm.c:2224

Only the `gc.c` spots look relevant, all deal with threads, and two (lines 572 and 739) deal with not having a ruby thread.

Re: [TheCodeShop] Cry for help regarding with the WDM gem issue.

jarm...@gmail.com (Jarmo Pertman) — Sun, 20 Jan 2013 15:36:27 UT

I have not run it with GDB - last time i tried to get it running on my
machine i was not successful :(

However, as i wrote on the GitHub issue, i did localize the problem by
commenting out C-code and then recompiling. In that way i was able to find
out the offending code which calls rb_thread_blocking_region and causes

Re: [TheCodeShop] Cry for help regarding with the WDM gem issue.

luislav...@gmail.com (Luis Lavena) — Sun, 20 Jan 2013 15:16:11 UT

Actually rubyinstaller list might have a higher number of Windows
developers, just saying ;-)

Have you run the code that generates this exception with GDB? Do you
have a backtrace when that exception was triggered to determine who
called rb_thread_blocking_region?

Did you compile the gem or was a pre-compiled binary?

Re: [TheCodeShop] Cry for help regarding with the WDM gem issue.

jarm...@gmail.com (Jarmo Pertman) — Sat, 19 Jan 2013 19:25:02 UT

Reproducilibity is always a good thing. Thank you, Jon!
Jarmo

Re: [TheCodeShop] Cry for help regarding with the WDM gem issue.

jon.for...@gmail.com (Jon) — Sat, 19 Jan 2013 17:15:12 UT

Quick test and haven't looked into the C yet.

I can repro using your sample code on win7 32bit (4GB) with ruby p374 (built with mingw-32-4.6.2) and wdm-0.0.3.gem

C:\Users\Jon\Documents\RubyDev \sandbox>ruby wdm_crash.rb
1: C:/Users/Jon/Documents/RubyDev /sandbox/foobar
...
120: C:/Users/Jon/Documents/RubyDev /sandbox/foobar