win: Only retry in UseHandler() loop on ERROR_PIPE_BUSY

win: Only retry in UseHandler() loop on ERROR_PIPE_BUSY

This is better because now end_to_end_test.py fails immediately with

[1180:9020:20151106,145204.830:ERROR registration_protocol_win.cc:39] CreateFile: The system cannot find the file specified.  (0x2)

R=mark@chromium.org
BUG=crashpad:75
Committed: https://chromium.googlesource.com/crashpad/crashpad/+/ff274507dc30cb3b4273a04cd73f3ab5240d02e7

[Mark Mentovai, 2015-11-06 23:03:34]

It looks like you and I collided, but I think our changes can complement each
other.

https://codereview.chromium.org/1409693011/diff/1/util/win/registration_proto...
File util/win/registration_protocol_win.cc (right):

https://codereview.chromium.org/1409693011/diff/1/util/win/registration_proto...
util/win/registration_protocol_win.cc:43: if (!WaitNamedPipe(pipe_name.c_str(),
1000)) {
We don’t want to do this at all on the last try.

https://codereview.chromium.org/1409693011/diff/1/util/win/registration_proto...
util/win/registration_protocol_win.cc:44: PLOG(ERROR) << "WaitNamedPipe";
If GetLastError() is ERROR_SEM_TIMEOUT, we should probably keep retrying.

[scottmg, 2015-11-06 23:10:59]

How about like this?

https://codereview.chromium.org/1409693011/diff/1/util/win/registration_proto...
File util/win/registration_protocol_win.cc (right):

https://codereview.chromium.org/1409693011/diff/1/util/win/registration_proto...
util/win/registration_protocol_win.cc:43: if (!WaitNamedPipe(pipe_name.c_str(),
1000)) {
On 2015/11/06 23:03:34, Mark Mentovai wrote:
> We don’t want to do this at all on the last try.

Done.

https://codereview.chromium.org/1409693011/diff/1/util/win/registration_proto...
util/win/registration_protocol_win.cc:44: PLOG(ERROR) << "WaitNamedPipe";
On 2015/11/06 23:03:34, Mark Mentovai wrote:
> If GetLastError() is ERROR_SEM_TIMEOUT, we should probably keep retrying.

Done.

[Mark Mentovai, 2015-11-06 23:32:35]

LGTM

https://codereview.chromium.org/1409693011/diff/40001/snapshot/win/end_to_end...
File snapshot/win/end_to_end_test.py (right):

https://codereview.chromium.org/1409693011/diff/40001/snapshot/win/end_to_end...
snapshot/win/end_to_end_test.py:107: print 'Waiting for crashpad_handler to be
ready...'
Only print this the first time. If crashpad_handler never gets ready, this will
spam output.

https://codereview.chromium.org/1409693011/diff/40001/util/win/registration_p...
File util/win/registration_protocol_win.cc (right):

https://codereview.chromium.org/1409693011/diff/40001/util/win/registration_p...
util/win/registration_protocol_win.cc:38: if (tries++ == 5 || GetLastError() !=
ERROR_PIPE_BUSY) {
We want ++tries to be equivalent to the previous behavior.

https://codereview.chromium.org/1409693011/diff/40001/util/win/registration_p...
util/win/registration_protocol_win.cc:72: LOG(ERROR) << "TransactNamedPipe read
incorrect number of bytes";
Can you grab the updated LOG() line from
https://codereview.chromium.org/1415453012/, which is otherwise obsoleted by
this change?

[scottmg, 2015-11-06 23:53:40]

https://codereview.chromium.org/1409693011/diff/40001/snapshot/win/end_to_end...
File snapshot/win/end_to_end_test.py (right):

https://codereview.chromium.org/1409693011/diff/40001/snapshot/win/end_to_end...
snapshot/win/end_to_end_test.py:107: print 'Waiting for crashpad_handler to be
ready...'
On 2015/11/06 23:32:35, Mark Mentovai wrote:
> Only print this the first time. If crashpad_handler never gets ready, this
will
> spam output.

Done.

https://codereview.chromium.org/1409693011/diff/40001/util/win/registration_p...
File util/win/registration_protocol_win.cc (right):

https://codereview.chromium.org/1409693011/diff/40001/util/win/registration_p...
util/win/registration_protocol_win.cc:38: if (tries++ == 5 || GetLastError() !=
ERROR_PIPE_BUSY) {
On 2015/11/06 23:32:35, Mark Mentovai wrote:
> We want ++tries to be equivalent to the previous behavior.

Done.

https://codereview.chromium.org/1409693011/diff/40001/util/win/registration_p...
util/win/registration_protocol_win.cc:72: LOG(ERROR) << "TransactNamedPipe read
incorrect number of bytes";
On 2015/11/06 23:32:35, Mark Mentovai wrote:
> Can you grab the updated LOG() line from
> https://codereview.chromium.org/1415453012/, which is otherwise obsoleted by
> this change?

Done.

[scottmg, 2015-11-06 23:54:55]

Committed patchset #4 (id:60001) manually as
ff274507dc30cb3b4273a04cd73f3ab5240d02e7 (presubmit successful).