Re: [Diffutils-devel] diffutils-3.6.17-d5ba on HP-UX

[Jim Meyering]

On Sun, Dec 30, 2018 at 10:26 AM Bruno Haible <address@hidden> wrote:
...
> The test scenario I used on Linux and HP-UX is:
> - Preparation:
>   $ cd tests
>   $ echo a > a
> - Command:
>   $ ../src/diff -N - a <&-
>
> Result on HP-UX:
> $ ../src/diff -N - a <&-
> diff: -: Bad file number
>
> Result on Linux:
> $ ../src/diff -N - a <&-
> 0a1
> > a
...
>
> On HP-UX:
>
> $ tusc ../src/diff -N - a <&-
> execve("../src/diff", 0x87fffffffffffb08, 0x87fffffffffffb30) = 0 [64-bit]
> ...
> ioctl(0, TCGETA, 0x87fffffffffff3e0) ..................... ERR#25 ENOTTY
> fstat(0, 0x87fffffffffff420) ............................. = 0
...
>
> On Linux:
>
> $ strace ../src/diff -N - a <&-
> execve("../src/diff", ["../src/diff", "-N", "-", "a"], [/* 78 vars */]) = 0
> ...
> ioctl(0, TCGETS, 0x7ffd75318810)        = -1 EBADF (Bad file descriptor)
> fstat(0, 0x7ffd75318940)                = -1 EBADF (Bad file descriptor)
...
>
> The crucial difference is the open("a", O_RDONLY...) call, which is meant

Hi Bruno,
Thanks for all of that analysis.

The crucial difference seems earlier than that. HP-UX's fstat
erroneously returns 0 for the invalid (closed) stdin file descriptor,
while Linux's returns EBADF.

GNU diff explicitly handles the case of a closed file descriptor (note
the test for errno == EBADF in diff.c). That test cannot do its job
when fstat is not POSIX-conforming. NEWS from diffutils-3.3 says this:

  --new-file (-N) and --unidirectional-new-file now allow comparisons to "-".
  A standard input that's closed acts like a nonexistent file.

One way to work around that HP-UX flaw is to make gnulib's fstat
replacement return EBADF when the return value would otherwise be 0,
yet fcntl (fd, F_GETFD) returns -1.