csplit: seq 10 | csplit - 1 silently ignores EIO (tests/csplit/csplit-io-err.sh?)

[oech3]

As similar with #9832 ,

$ run0 dmsetup create ioerr --table "0 8 error"
$ chmod 777 /dev/mapper/ioerr
$ ln -sf /dev/mapper/ioerr xx01
$ seq 10 | uu-csplit - 1 # success
$ file xx01
xx01: symbolic link to /dev/mapper/ioerr
$ seq 10 | gnu-csplit - 1
0
csplit: write error for 'xx01': Input/output error
$ file xx01
xx01: cannot open `xx01' (No such file or directory)

[oech3]

It seems this is not csplit-io-err.sh too. @ChrisDryden Can you remove LD_PRELOAD?

[ChrisDryden]

Are you saying to rewrite the test to not use LD_PRELOAD? From initial investigation I think the easier adapting is to have a wrapper for fwrite. I have modified the test to work with write but it appears there is still an underlying issue that needs to be fixed.

[oech3]

Are you saying to rewrite the test to not use LD_PRELOAD?

Yes. I want to avoid having patches/sed hacks at downstream if possible (or keep upstreamable form).

[ChrisDryden]

I'll see if they're open to the idea of changing that test to be a symlink to /dev/full since that would give the error code. Have you gone through the process of proposing changing the tests, I have an example new test that would cover the same cases but I'm not familiar with how open they are to update the tests?

[ChrisDryden]

  #!/bin/sh
  # Ensure we handle i/o errors correctly in csplit

  # Copyright (C) 2015-2025 Free Software Foundation, Inc.

  # This program is free software: you can redistribute it and/or modify
  # it under the terms of the GNU General Public License as published by
  # the Free Software Foundation, either version 3 of the License, or
  # (at your option) any later version.

  # This program is distributed in the hope that it will be useful,
  # but WITHOUT ANY WARRANTY; without even the implied warranty of
  # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  # GNU General Public License for more details.

  # You should have received a copy of the GNU General Public License
  # along with this program.  If not, see <https://www.gnu.org/licenses/>.

  . "${srcdir=.}/tests/init.sh";
  print_ver_ csplit

  if ! test -w /dev/full || ! test -c /dev/full; then
    skip_ '/dev/full is required'
  fi

  # Ensure error messages are in English
  LC_ALL=C
  export LC_ALL

  # Get the wording of the OS-dependent ENOSPC message
  returns_ 1 seq 1 >/dev/full 2>msgt || framework_failure_
  sed 's/seq: write error: //' msgt > msg || framework_failure_

  # Create the expected error message
  { printf "%s" "csplit: write error for 'xx01': " ; cat msg ; } > exp \
    || framework_failure_

  # Create symlink so xx01 writes to /dev/full (triggers ENOSPC)
  ln -s /dev/full xx01 || framework_failure_

  # Split the input - xx00 succeeds (empty), xx01 fails
  # the 'csplit' command should fail with exit code 1
  seq 10 | returns_ 1 csplit - 1 4 2>out || fail=1

  # Ensure we got the expected error message
  compare exp out || fail=1

  Exit $fail

[oech3]

I think csplit: write error for 'xx01' is difficult to repro by symlink.

[oech3]

#9832 (comment)

[ChrisDryden]

I think its only because we have an issue with our implementation, let me publish my draft branch and it works with the symlink test too

[oech3]

Can we see write error with gnu/csplit with symlink instead of no spce or I/O error?

[ChrisDryden]

I think #9855 addresses that? Let me create the commands you can run to validate that this matches the gnu implementation

[oech3]

Fixed. Thankyou. It would be nice to use the word "Fixes ISSUE_TAG" at next PR.