seq: returning exit code 1 on broken pipe, fixes seq-epipe

[ChrisDryden]

Work in progress for me trying to understand the whole EPIPE situation, on one hand the GNU implementation is supposed to clearly return a 1 in this case, but when the ignore sigpipe is set it should return 0 and we don't have that in any of the utilities. I think the plan could be set this for now and when the library that all of the utilities use for ignoring the sig-pipe is made we add it here.

[github-actions[bot]]

GNU testsuite comparison:

Congrats! The gnu test tests/seq/seq-epipe is no longer failing!

[pixelb]

There must have been a misunderstanding in the original code before this change. GNU seq never returns 0 upon pipe error. The shell implicitly handles SIGPIPE result from any command, which I guess is the source of the misunderstanding. For the record seq (or any properly behaving unix command) returns failure upon any pipe error. See:

$ env --ignore-signal=PIPE seq inf | :
seq: write error: Broken pipe
$ echo ${PIPESTATUS[@]}
1 0

$ seq inf | :
$ echo ${PIPESTATUS[@]}
141 0

pipes being central to the unix model, this is a critical thing to get right, though it is an often misunderstood signal, as I've detailed at: https://www.pixelbeat.org/programming/sigpipe_handling.html

[ChrisDryden]

Started working on a draft that is way more comprehensive and is able to pass all of the GNU pipe signal related tests based on @mattsu2020 's handler https://github.com/uutils/coreutils/pull/9646 will be separating the shared library and the seq fixes into a separate PR