Interview with Valgrind Author Julian Seward (KDE.News) [LWN.net]

Excellent Tool

Posted Feb 21, 2006 23:24 UTC (Tue) by fyodor (guest, #3481) [Link]

Valgrind is great! It has helped me track down numerous memory leaks and other subtle errors in the Nmap Security Scanner. It is one of those programs (kinda like Nmap, heh) that often gives you a ton of data to mine through to find what you are looking for. But it is still better at identifying problems than any of the free alternatives I've tried.

The article is titled "Interview with Valgrind Author Julian Seward", but he may be better known as the author of bzip2. I'd like to see a second interview that focuses on his data compression thoughts instead.

-Fyodor

Excellent Tool

Posted Feb 22, 2006 11:43 UTC (Wed) by nix (subscriber, #2304) [Link]

In my experience it's better than proprietary stuff like Purify too. It's been a major factor pushing us from SPARC/Solaris to (x86/x86-64)/Linux on development boxes at work.

Excellent Tool

Posted Feb 22, 2006 22:59 UTC (Wed) by jreiser (subscriber, #11027) [Link]

Please share the reasons that you consider Valgrind/memcheck to be better than Purify.

Excellent Tool

Posted Feb 22, 2006 23:55 UTC (Wed) by xoddam (subscriber, #2322) [Link]

I've never had an employer consider a project I'm working on important
enough to provide expensive tools like Purify -- so my experience with
that tool is mostly second-hand. I can do with Valgrind for free
everything I ever thought I wanted Purify for.

Purify does provide a little more -- performance is somewhat better, and
it has support for Java and Visual Basic bytecode, for example -- but
unless you desperately need those extra features, I don't think they
justify the cost.

Valgrind v. Purify

Posted Feb 23, 2006 5:43 UTC (Thu) by brugolsky (subscriber, #28) [Link]

Purify is an excellent developer tool -- and it's pretty fast. I used to use it on Solaris/SPARC. Had less luck getting it to work on Fedora; it really wanted the RHEL glibc. But it's major drawback was (not sure if this is still the case) that it required that one use the purify wrapper to link the application.

Valgrind is an excellent developer tool. It is also an end-user tool. Say I, Joe User, hit a bug in ... Evolution. Now, if I'm using Purify (at least the last version that I used), I can go grab the source, recompile and relink with Purify, and attempt to reproduce the bug. Or I can instead grab the debuginfo rpms for Evolution and its libraries, install them, and run the process using Valgrind. Next stop, Bugzilla, where I can post the precise error and backtrace.

Valgrind v. Purify

Posted Feb 24, 2006 4:15 UTC (Fri) by JoeBuck (subscriber, #2330) [Link]

Purify only requires that you relink; it can deal with object code you don't have the source to. Nevertheless, that's still a lot less convenient than valgrind, which can run any binary as long as it is not statically linked.

Excellent Tool

Posted Feb 23, 2006 16:13 UTC (Thu) by zooko (subscriber, #2589) [Link]

Valgrind tracks the validity status of every bit of data. This is dramatically better than Purify used to do a few years ago. Possibly the most important way that this is dramatically better is that it dramatically reduces the occurence of false alarms.

http://valgrind.org/docs/manual/mc-manual.html#mc-manual....

http://valgrind.org/docs/pubs.html

Perhaps modern Purify does likewise. I don't know.

Everyone who needs to debug native code on Linux on x86, x86-64 or PowerPC, ought to try valgrind. It is easy to install it and run it and see the results, and it has a very high ratio of true alarms to false alarms. It's worth it!