On Thu, 30 Nov 1995, Bruce Perens wrote: > Note that "ls" lists the length of /proc/filesystems > as "0", but read() will get more than 0 data. Less doesn't know that > /proc files are special in this way, and treats them as a 0-length file. > > You could argue that this is a bug in the semantics of the proc filesystem. > For now, I think we should fix "less" to not abandon any attempt to read a > file when stat() says it's 0-length.
I do so argue. However, I've uploaded an updated a.out less package with a hacked-in workaround for this. Would the kernel package maintainer please report this as a bug to the kernel group? Both stat() and lseek() seem to be broken for proc filesystem files.
