From: Clifford Heath Date: 2011-06-28T12:05:07+09:00 Subject: [ruby-core:37600] Re: CI? I see quite a few segfault bugs being reported. Does anyone use Purify, valgrind, or any such tools? I've used Purify extensively in the past, and regard it as an essential tool for any C or C++ development. Such tools should be *mandatory* for any widely-used code-bases. Memory corruption bugs are often innocuous, causing no effects until a slight change, that can be years after, causes random and unrelated crashes to appear seemingly out of the blue. Even just re-linking with an updated C library, adding a function that affects memory layout (so a bug in *another* function shows up), and basically the tiniest changes, can trigger a fault to show itself. The only way to effectively prevent such bugs is a tool like Purify, which isolates the source of a potential error, possibly long before the subsequent crash. Purify is much better than valgrind, by the way. It rewrites your object code (including all system shared libraries) such that every memory reference is checked; no bounds overflows, no read-before-set errors, etc. An incredible and under-utilised tool... Clifford Heath. On 28/06/2011, at 7:45 AM, NARUSE, Yui wrote: > (2011/06/28 6:28), Ryan Davis wrote: >> Is this an official CI for ruby? > > It is mine. > It is the result of chkbuild on FreeBSD 8.2 x86_64. > chkbuild is a CI by akr. > http://cvs.m17n.org/viewcvs/ruby/chkbuild/ > > >> http://59.106.172.211/~chkbuild/ruby-trunk/summary.html >> >> If so... how does one read this?? I can't make sense of it. > > For examle, > 20101011T060001Z ruby 1.9.3dev (2010-10-11 trunk 29438) [x64- > freebsd8.1] (fbsd81) 2730W 1F0E1S rubyspec:23F1E (diff:test-all) > > 20101011T060001Z is the time chkdbuild runs from; the link is for > the result. > ruby 1.9.3dev (2010-10-11 trunk 29438) [x64-freebsd8.1] is `ruby -v`. > (fbsd81) is the name of this CI. > 2730W is the number of warnings of the result. > 1F0E1S is the result of test-all; 1 Fail 0 Error 1 Skip. > rubyspec:23F1E is the result of test-rubyspec; 23 Failure 1 Error. > (diff:test-all) is the link to the diff between this result and the > previous result. > > Anyway, summary.html is too big; I recommend to see recent.html. > > We should keep 0F0E, but in current situation on the environment > 1F0E46S rubyspec:1F1E is the best result. > If it is increased, it means someone breaks something. > > -- > NARUSE, Yui >