Re:Stripping kernel/uboot source to 10% for code reading
by unoti on Wednesday May 08, 2013 @12:37PM
(#5675190)
> git ls-tree HEAD -r |pn 4|xargs touch
What's pn? I've never heard of it, I don't have it, and simple searches don't tell me anything about it. It looks like it might be a simple replacement for awk argument parsing, but I'm unsure...
Re:Stripping kernel/uboot source to 10% for code reading
by adamtj on Wednesday May 08, 2013 @03:00PM
(#5676145)
> touch all source code files (this step is mandatory, or else the atime won't update correctly, I don't know why, but guess it's an optimization).
On a filesystem mounted with the "relatime" option, it's necessary to touch each file so that the next access is guaranteed to bump the atime.
It used to be common to update the access time every time a file was accessed. That turned every read into a read+write, which is expensive. Linux filesystems are now commonly mounted with either the "relatime" or "noatime" options. "noatime" does what you'd expect. "relatime" is a compromise. It updates the atime on access only if it's the same as the mtime -- in other words, only once after a write. I think some mail or news readers consider a message to be read only if atime > mtime. They won't work right with "noatime".
Re:Stripping kernel/uboot source to 10% for code reading
by baohaojun on Thursday May 09, 2013 @01:50AM
(#5679015)
Thanks for the info, it must be the relatime that I saw this behavior. Were it noatime, this idea won't have worked at all (I think I have seen noatime used on embedded devices).
Re:Stripping kernel/uboot source to 10% for code reading
by jholman on Wednesday May 08, 2013 @05:37PM
(#5677122)
That was a great talk, too great for just an upvote. Thanks!
To others considering watching it, the speaker is willing to indulge himself in a bit of self-absorbed wankery, but he gets most of that out of the way in the first 10 minutes, and gets down to the useful business of "here's an example problem, and here's a solution".
What's pn? I've never heard of it, I don't have it, and simple searches don't tell me anything about it. It looks like it might be a simple replacement for awk argument parsing, but I'm unsure...
http://packages.debian.org/search?searchon=contents&keyw...
On a filesystem mounted with the "relatime" option, it's necessary to touch each file so that the next access is guaranteed to bump the atime.
It used to be common to update the access time every time a file was accessed. That turned every read into a read+write, which is expensive. Linux filesystems are now commonly mounted with either the "relatime" or "noatime" options. "noatime" does what you'd expect. "relatime" is a compromise. It updates the atime on access only if it's the same as the mtime -- in other words, only once after a write. I think some mail or news readers consider a message to be read only if atime > mtime. They won't work right with "noatime".
http://serverfault.com/questions/47466/drawbacks-of-mounting...
If so, then watch this talk: http://www.youtube.com/watch?v=TgmA48fILq8
Now you will prefer Solaris/BSD to Linux :)
To others considering watching it, the speaker is willing to indulge himself in a bit of self-absorbed wankery, but he gets most of that out of the way in the first 10 minutes, and gets down to the useful business of "here's an example problem, and here's a solution".
Edit: I'm listening to it now and it's excellent.