STDIN inheritance during exec and difference between getc and sysread

Ruby
#1

Hi all!

I’ve accidentally found that “getc” and “sysread” behave differently
after “exec”
getc does not get remaining data from the STDIN but sysread does it

[email protected] ~ $
[email protected] ~ $
[email protected] ~ $ echo -n ab | ruby -e “p STDIN.getc; exec ‘ruby’,
‘-e’, ‘p STDIN.getc’”
“a”
nil
[email protected] ~ $
[email protected] ~ $ echo -n ab | ruby -e “p STDIN.sysread(1); exec
‘ruby’, ‘-e’, ‘p STDIN.sysread(1)’”
“a”
“b”
[email protected] ~ $
[email protected] ~ $
[email protected] ~ $

why so? where is “b” in the first case?

#2

Dmitry B. [email protected] wrote:

“a”
why so? where is “b” in the first case?
getc uses Ruby I/O buffering in user space, so the process read
more bytes than it needed the first time getc was called.

Buffered I/O leads to surprises like this behavior across fork/exec.