Re:Examples of errors detected in various open-source projects
by mynegation on Wednesday November 02, 2011 @10:53AM
(#3186997)
I worked for static analysis company.
I am happy to see this kind of analyses popping up here and there. As computers become more powerful, you can do more sophisticated checks.
One of the biggest problems in static analysis is false positives - reports about potential problems which are not relevant for one reason or another, and false positive rate has huge impact on ROI.
Unfortunately, aside from an interface element for triage, article does not mention or link to the article that describes how they are handling false-positives problem in the analysis.
Re:Examples of errors detected in various open-source projects
by dexen on Wednesday November 02, 2011 @11:19AM
(#3187149)
The article does not mention whether the bugs have been reported and rectified in subsequent releases, either.
Since at least some of the bugs found are segfault/data-corruption/wrong-query-results level ones[†], one would assume they'd get found out quite soon.
For me, when picking software for a new project, an important metric is how fast bugs get fixed upstream.
--
[†] I just love how `ones' is a perfectly reasonable expression in english. `Would that it were so in our programming languages', with apologies to Alan Perlis ;-)
Re:Examples of errors detected in various open-source projects
by pavel_lishin on Wednesday November 02, 2011 @10:40AM
(#3186903)
Yes, but not necessarily easily integrated into the next release. You can fix it for yourself, but if the project isn't very well maintained, or if the maintainer rejects your patch, etc., etc...
Re:Examples of errors detected in various open-source projects
by dexen on Wednesday November 02, 2011 @11:27AM
(#3187195)
That's why some distros roll packages with bugfix patches applied even before upstream maintainers include those in official releases.
Even better, if a particular bug is a show-stopper at your organization, you can build fixed package and deploy it to your local package repository on a short notice; machines can then obtain the update automagically.
Re:Examples of errors detected in various open-source projects
by danso on Wednesday November 02, 2011 @09:20AM
(#3186496)
Fascinating...on one hand, it's disheartening to see the vast array of errors that can go undetected and how easy they are missed even in retrospect. On the other, it's nice to know I'm not the only one :)
...cause the article is marked as `deleted' now :-(
I am happy to see this kind of analyses popping up here and there. As computers become more powerful, you can do more sophisticated checks.
One of the biggest problems in static analysis is false positives - reports about potential problems which are not relevant for one reason or another, and false positive rate has huge impact on ROI.
Unfortunately, aside from an interface element for triage, article does not mention or link to the article that describes how they are handling false-positives problem in the analysis.
Since at least some of the bugs found are segfault/data-corruption/wrong-query-results level ones[†], one would assume they'd get found out quite soon.
For me, when picking software for a new project, an important metric is how fast bugs get fixed upstream.
--
[†] I just love how `ones' is a perfectly reasonable expression in english. `Would that it were so in our programming languages', with apologies to Alan Perlis ;-)
Even better, if a particular bug is a show-stopper at your organization, you can build fixed package and deploy it to your local package repository on a short notice; machines can then obtain the update automagically.
http://cacm.acm.org/magazines/2010/2/69354-a-few-billion-lin...
Though kinda interesting anyway if a little blandly written.