Comments on: Worst lock acquisition code ever. http://kenkinder.com/2008/11/24/worst-lock-acquisition-code-ever/ Ken Kinder's personal website Wed, 02 Jun 2010 07:46:06 +0000 hourly 1 By: Jonathan Briggs http://kenkinder.com/2008/11/24/worst-lock-acquisition-code-ever/comment-page-1/#comment-472 Jonathan Briggs Mon, 08 Dec 2008 07:58:06 +0000 /?p=82#comment-472 Liferea, a Gnome RSS reader. It isn't showing an error today so I can't see the exact problem. Liferea, a Gnome RSS reader. It isn’t showing an error today so I can’t see the exact problem.

]]> By: Ken Kinder http://kenkinder.com/2008/11/24/worst-lock-acquisition-code-ever/comment-page-1/#comment-461 Ken Kinder Wed, 26 Nov 2008 18:47:55 +0000 /?p=82#comment-461 Oh, and yeah, I thought I would start blogging again. I gave up on writing my own web framework and am now using WordPress. It's really a very impressive little package, especially when I throw on some useful plugins. Even though I'm loathed to use anything in PHP. :) Oh, and yeah, I thought I would start blogging again. I gave up on writing my own web framework and am now using WordPress. It’s really a very impressive little package, especially when I throw on some useful plugins.

Even though I’m loathed to use anything in PHP. :)

]]> By: Ken Kinder http://kenkinder.com/2008/11/24/worst-lock-acquisition-code-ever/comment-page-1/#comment-460 Ken Kinder Wed, 26 Nov 2008 18:47:13 +0000 /?p=82#comment-460 The one here? http://kenkinder.com/comments/feed/ It looks escaped correctly to me and Firefox parses it. What RSS reader are you using? The one here?

http://kenkinder.com/comments/feed/

It looks escaped correctly to me and Firefox parses it. What RSS reader are you using?

]]> By: Ken Kinder http://kenkinder.com/2008/11/24/worst-lock-acquisition-code-ever/comment-page-1/#comment-459 Ken Kinder Wed, 26 Nov 2008 18:46:32 +0000 /?p=82#comment-459 Yeah, sometimes it's astonishing how resistant Open Source developers are to bug fixes. If you see my thread on CherryPy, it has yet to be answered. And since I posted it, a new version has come out, without a fix. Now CherryPy's session code is rendering 500 errors when the user has an invalid cookie, but writes nothing in the log file. The whole thing is making me re-evaluate my choice of CherryPy. Yeah, sometimes it’s astonishing how resistant Open Source developers are to bug fixes. If you see my thread on CherryPy, it has yet to be answered. And since I posted it, a new version has come out, without a fix. Now CherryPy’s session code is rendering 500 errors when the user has an invalid cookie, but writes nothing in the log file. The whole thing is making me re-evaluate my choice of CherryPy.

]]> By: Jonathan Briggs http://kenkinder.com/2008/11/24/worst-lock-acquisition-code-ever/comment-page-1/#comment-458 Jonathan Briggs Wed, 26 Nov 2008 18:06:29 +0000 /?p=82#comment-458 Also your blog has a bug in the comments feed that I am seeing in the modem chat strings Aaron wrote. It isn't escaping the ampersands correctly, creating invalid XML. Also your blog has a bug in the comments feed that I am seeing in the modem chat strings Aaron wrote. It isn’t escaping the ampersands correctly, creating invalid XML.

]]> By: Jonathan Briggs http://kenkinder.com/2008/11/24/worst-lock-acquisition-code-ever/comment-page-1/#comment-457 Jonathan Briggs Wed, 26 Nov 2008 18:03:00 +0000 /?p=82#comment-457 Hi Ken, glad to see you back blogging. I was surprised when my RSS reader popped up with new entries from you. It's been a long time! This lock method is pretty bad I agree. I've seen it in C code as well, although C programmers seem a bit more likely to check for a specific error code. The worst/most annoying file locking bug I've seen is in Evolution, the email program. Their mbox update code locked the original mbox file while writing a .tmp copy with the updated contents. It then used rename() to replace the original with the updated copy. At this point, the file lock is being held on the origninal file which is now DELETED! But then their code continued on as if the mbox file was still locked. After I figured that one out and filed a bug report they REFUSED to apply my patch to fix it because they claimed it didn't matter, then got abusive when I asked why they were bothing to lock the file in the first place. sigh. Hi Ken, glad to see you back blogging. I was surprised when my RSS reader popped up with new entries from you. It’s been a long time!

This lock method is pretty bad I agree. I’ve seen it in C code as well, although C programmers seem a bit more likely to check for a specific error code.

The worst/most annoying file locking bug I’ve seen is in Evolution, the email program. Their mbox update code locked the original mbox file while writing a .tmp copy with the updated contents. It then used rename() to replace the original with the updated copy. At this point, the file lock is being held on the origninal file which is now DELETED! But then their code continued on as if the mbox file was still locked.

After I figured that one out and filed a bug report they REFUSED to apply my patch to fix it because they claimed it didn’t matter, then got abusive when I asked why they were bothing to lock the file in the first place.

sigh.

]]>