Re: [platform-testers] new snapshot available: grep-3.4-almost.19-ff30

[Paul Eggert]

On 9/20/20 5:07 AM, Jeffrey Walton wrote:
> I'm retesting Alpine 3.10 with grep-3.4-almost.31-fe6c. I'm seeing a
> few squawks:
>
> FAIL: grep-dev-null-out

I suspect that the 'splice' system call is not working correctly on Alpine 3.10, 
when copying from a pipe to /dev/null. To test this hypotheses, try something 
like this:

export LC_ALL=C
mkfifo fifo
awk 'BEGIN {while (1) print "x"}' </dev/null >fifo &
timeout 1 strace -o tr src/grep x <fifo >/dev/null
echo $?

On my platform (Fedora 31 x86-64), this outputs "124" and "tr" ends like this:

fstat(1, {st_mode=S_IFCHR|0666, st_rdev=makedev(0x1, 0x3), ...}) = 0
stat("/dev/null", {st_mode=S_IFCHR|0666, st_rdev=makedev(0x1, 0x3), ...}) = 0
fstat(0, {st_mode=S_IFIFO|0664, st_size=0, ...}) = 0
lseek(0, 0, SEEK_CUR)                   = -1 ESPIPE (Illegal seek)
read(0, "x\nx\nx\nx\nx\nx\nx\nx\nx\nx\nx\nx\nx\nx\nx\nx\n"..., 98304) = 65536
splice(0, NULL, 1, NULL, 98304, SPLICE_F_MOVE) = 4096
splice(0, NULL, 1, NULL, 98304, SPLICE_F_MOVE) = 4096
... [many more duplicates like this]
splice(0, NULL, 1, NULL, 98304, SPLICE_F_MOVE) = 4096
splice(0, NULL, 1, NULL, 98304, SPLICE_F_MOVE) = ? ERESTARTSYS (To be restarted 
if SA_RESTART is set)
--- SIGTERM {si_signo=SIGTERM, si_code=SI_USER, si_pid=1615171, si_uid=1000} --\
-
+++ killed by SIGTERM +++


What does it do on Alpine 3.10? If there's a way to determine that 'splice' may 
be buggy, perhaps we can teach 'grep' to ignore 'splice' on that platform.