We should not try to read from /dev/null. While some distributions define this behavior well, Darwin, notably, does not. Use PIPE instead.
<rdar://problem/88836868>
Stumbled into this when investigating https://bugs.webkit.org/show_bug.cgi?id=230513. This change isn't implicated in that bug, but we should still fix this.
Created attachment 451749 [details] Patch
Comment on attachment 451749 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=451749&action=review > Tools/ChangeLog:9 > + While writing to /dev/null is well defined, reading from it is not. Do not I didn't know this. What is the source of this information? That said, the patch removes more code than it adds, and webkitpy tests pass, so it's good.
Man page for 'null': NAME null – the null device DESCRIPTION The null device accepts and reads data as any ordinary (and willing) file - but throws it away. The length of the null device is always zero. FILES /dev/null HISTORY A null device appeared in Version 7 AT&T UNIX. Notably, says nothing about what happens if you try and read from this file.
Committed r289703 (247188@main): <https://commits.webkit.org/247188@main> All reviewed patches have been landed. Closing bug and clearing flags on attachment 451749 [details].
Reading from /dev/null is standardized, https://pubs.opengroup.org/onlinepubs/9699919799/