There are times on the Mac when pipe names of 0 length may occur. Let the caller handle

There are times on the Mac when pipe names of 0 length may occur. Let the caller handle
the error instead of aborting here.

BUG=75518
TEST=BUILD

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=98225

[dmac, 2011-08-24 22:20:04]

If one of you could PTAL I'd appreciate it. Should be pretty quick.

[agl, 2011-08-24 22:23:03]

LGTM

[jeremy, 2011-08-25 07:16:51]

LGTM

http://codereview.chromium.org/7743001/diff/1/ipc/ipc_channel_posix_unittest.cc
File ipc/ipc_channel_posix_unittest.cc (right):

http://codereview.chromium.org/7743001/diff/1/ipc/ipc_channel_posix_unittest....
ipc/ipc_channel_posix_unittest.cc:298: "leading-edge_processes");
_with_a_cherry_on_top :o)

[Mark Mentovai, 2011-08-25 13:11:02]

LG otherwise

http://codereview.chromium.org/7743001/diff/1/ipc/ipc_channel_posix_unittest.cc
File ipc/ipc_channel_posix_unittest.cc (right):

http://codereview.chromium.org/7743001/diff/1/ipc/ipc_channel_posix_unittest....
ipc/ipc_channel_posix_unittest.cc:291: IPC::ChannelHandle
handle2("This_is_a_very_long_name_that_must_be_more_than"
Do “const char kTooLongName[] =
This_is_a_very_long_name_that_must_be_more_than"…” and then
EXPECT_GE(strlen(kTooLongName), IPC::kMaxPipeNameLength).

Since we expect the buffer size on Linux to be 108 but cap it at 104 anyway (see
ipc_channel_posix.h), you may also want to EXPECT_LT(strlen(kTooLongName), 108)
to ensure that we’re testing the right thing on Linux to. If you don’t like
hardcoding 108, you can do:

  if (sizeof(((sockaddr_un*)0)->sun_path) > IPC::kMaxPipeNameLength)
    EXPECT_LT(strlen(kTooLongName), sizeof(((sockaddr_un*)0)->sun_path));

http://codereview.chromium.org/7743001/diff/1/ipc/ipc_channel_posix_unittest....
ipc/ipc_channel_posix_unittest.cc:300: ASSERT_FALSE(channel2.Connect());
Note that the suggested size checks could be ASSERT instead of EXPECT, but I
think that the Connect() checks should be EXPECT instead of ASSERT.