[PATCH] t/cmd_ipc: allow extra errors and add diagnostics

[Eric Wong <e@80x24.org>]

Konstantin Ryabitsev <konstantin@linuxfoundation.org> wrote:
>     t/cmd_ipc.t .................. 29/?
>     #   Failed test 'got EMSGSIZE'
>     #   at t/cmd_ipc.t line 108.
>     # Looks like you failed 1 test of 42.
>     t/cmd_ipc.t .................. Dubious, test returned 1 (wstat 256, 0x100)
>     Failed 1/42 subtests
>             (less 13 skipped subtests: 28 okay)

I think this is either caused by too much RAM and/or
/proc/sys/net/core/wmem_* being larger-than-expected
(both default to 212992 for me).

This fixes failures when I set both wmem_max and wmem_default
to 2129920 (10x its default value).

-------------8<-----------
Subject: [PATCH] t/cmd_ipc: allow extra errors and add diagnostics

Apparently, sendmsg can fail in less common ways when
network buffers are gigantic.  Add some diagnostics for
future failures, as well.

Link: https://public-inbox.org/meta/20210927124056.kj5okiefvs4ztk27@meerkat.local/
---
 t/cmd_ipc.t | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/t/cmd_ipc.t b/t/cmd_ipc.t
index c5e715a1cd8a..dd90fa2a63e4 100644
--- a/t/cmd_ipc.t
+++ b/t/cmd_ipc.t
@@ -85,7 +85,9 @@ my $do_test = sub { SKIP: {
 			$nsent += $n;
 			fail "sent 0 bytes" if $n == 0;
 		}
-		ok($!{EAGAIN}, "hit EAGAIN on send $desc");
+		ok($!{EAGAIN} || $!{ETOOMANYREFS},
+			"hit EAGAIN || ETOOMANYREFS on send $desc") or
+			diag "send failed with: $!";
 		ok($nsent > 0, 'sent some bytes');
 
 		socketpair($s1, $s2, AF_UNIX, $type, 0) or BAIL_OUT $!;
@@ -105,8 +107,9 @@ my $do_test = sub { SKIP: {
 				diag "sent $nr, retrying with more";
 				$nr += 2 * 1024 * 1024;
 			} else {
-				ok($!{EMSGSIZE}, 'got EMSGSIZE');
-				# diag "$nr bytes hits EMSGSIZE";
+				ok($!{EMSGSIZE} || $!{ENOBUFS},
+					'got EMSGSIZE or ENOBUFS') or
+					diag "$nr bytes fails with: $!";
 				last;
 			}
 		}