Hi Pierre-Jean,
> Well, as on Fedora I get "a b".
> I've looked at the arch packages, it doesn't contain any patches, just
> a site.tmac for man and mdoc. Here is the pkgbuild:
> http://repos.archlinux.org/wsvn/packages/groff/repos/core-i686/PKGBUILD
I'll try and have a dig around there.
> So, I need more investigation, looking at my own installation. I'm
> gonna do some tests and research and reply if I find something.
One more test I tried,
$ printf 'foo\n\n' |
> strace -fe trace=execve,read nroff <(printf 'a\n.rd\nb\n') 2> >(egrep
'execve|read\(0,') |
> grep .
a foo b
execve("/usr/bin/nroff", ["nroff", "/dev/fd/63"], [/* 55 vars */]) = 0
[pid 17907] execve("/usr/bin/locale", ["locale", "charmap"], [/* 55 vars
*/]) = 0
[pid 17908] execve("/usr/bin/groff", ["groff", "-mtty-char", "-Tutf8",
"/dev/fd/63"], [/* 56 vars */]) = 0
[pid 17909] execve("/usr/bin/troff", ["troff", "-mtty-char", "-Tutf8",
"/dev/fd/63"], [/* 57 vars */]) = 0
[pid 17910] execve("/usr/bin/grotty", ["grotty"], [/* 57 vars */]) = 0
[pid 17910] read(0, <unfinished ...>
[pid 17909] read(0, "foo\n\n", 1024) = 5
$
With the "a foo b" output I can see PID 17909 attempts to read from
standard input (file descriptor 0) and gets the intended data. And
17909 earlier did a execve(2) of troff which is as I'd expect; troff
is the intended recipient. grotty has an unfinished read, not sure what
that's about.
Perhaps file descriptor 0 has been closed by the time troff tries to
read(2) from it on your errant system?
Cheers,
Ralph.
