Merge lp:~doanac/lava-dispatcher/serial-delay into lp:lava-dispatcher

[Andy Doan]

this fixes the serial input issue with TC2 and may help other platforms as well.

tested on panda and tc2.

[Dave Pigott (dpigott)]

+1 in general. I'm a little concerned that removing the longer initial senddelay might introduce a regression, but that's only a gut reaction.

review: Approve

[Michael Hudson-Doyle (mwhudson)]

happy to see this land if it works :)

we'll need to remember this code if we ever send large amounts of data over serial (but lets try not to do that please)

[Antonio Terceiro (terceiro)]

On Sat, Dec 15, 2012 at 12:16:23AM -0000, Andy Doan wrote:
> === modified file 'lava_dispatcher/utils.py'
> --- lava_dispatcher/utils.py 2012-11-26 19:30:28 +0000
> +++ lava_dispatcher/utils.py 2012-12-15 00:15:28 +0000
> @@ -163,13 +163,23 @@
>
> class logging_spawn(pexpect.spawn):
>
> + def __init__(self, command, timeout=30, logfile=None):
> + pexpect.spawn.__init__(
> + self, command, timeout=timeout, logfile=logfile)
> +
> + # serial can be slow, races do funny things, so increase delay
> + self.delaybeforesend = 0.05
> +

Maybe we want this to be configurable instead of hardcoded?

--
Antonio Terceiro
Software Engineer - Linaro
http://www.linaro.org

[Dave Pigott (dpigott)]

Agree. That way we could also experimenting to get it down to a minimum.

Sent from my Aldis Lamp

On 17 Dec 2012, at 14:34, Antonio Terceiro <email address hidden> wrote:

> On Sat, Dec 15, 2012 at 12:16:23AM -0000, Andy Doan wrote:
>> === modified file 'lava_dispatcher/utils.py'
>> --- lava_dispatcher/utils.py 2012-11-26 19:30:28 +0000
>> +++ lava_dispatcher/utils.py 2012-12-15 00:15:28 +0000
>> @@ -163,13 +163,23 @@
>>
>> class logging_spawn(pexpect.spawn):
>>
>> + def __init__(self, command, timeout=30, logfile=None):
>> + pexpect.spawn.__init__(
>> + self, command, timeout=timeout, logfile=logfile)
>> +
>> + # serial can be slow, races do funny things, so increase delay
>> + self.delaybeforesend = 0.05
>> +
>
> Maybe we want this to be configurable instead of hardcoded?
>
> --
> Antonio Terceiro
> Software Engineer - Linaro
> http://www.linaro.org
>
> https://code.launchpad.net/~doanac/lava-dispatcher/serial-delay/+merge/140050
> You are reviewing the proposed merge of lp:~doanac/lava-dispatcher/serial-delay into lp:lava-dispatcher.

[Michael Hudson-Doyle (mwhudson)]

Antonio Terceiro <email address hidden> writes:

> On Sat, Dec 15, 2012 at 12:16:23AM -0000, Andy Doan wrote:
>> === modified file 'lava_dispatcher/utils.py'
>> --- lava_dispatcher/utils.py 2012-11-26 19:30:28 +0000
>> +++ lava_dispatcher/utils.py 2012-12-15 00:15:28 +0000
>> @@ -163,13 +163,23 @@
>>
>> class logging_spawn(pexpect.spawn):
>>
>> + def __init__(self, command, timeout=30, logfile=None):
>> + pexpect.spawn.__init__(
>> + self, command, timeout=timeout, logfile=logfile)
>> +
>> + # serial can be slow, races do funny things, so increase delay
>> + self.delaybeforesend = 0.05
>> +
>
> Maybe we want this to be configurable instead of hardcoded?

I'll review that branch really quickly!

=== modified file 'lava_dispatcher/device/master.py'
--- lava_dispatcher/device/master.py	2012-12-13 21:23:55 +0000
+++ lava_dispatcher/device/master.py	2012-12-15 00:15:28 +0000
@@ -338,8 +338,6 @@
         while retry_count < retry_limit:
             proc = logging_spawn(cmd, timeout=1200)
             proc.logfile_read = self.sio
-            #serial can be slow, races do funny things, so increase delay
-            proc.delaybeforesend = 1
             logging.info('Attempting to connect to device')
             match = proc.expect(patterns, timeout=10)
             result = results[match]

=== modified file 'lava_dispatcher/utils.py'
--- lava_dispatcher/utils.py	2012-11-26 19:30:28 +0000
+++ lava_dispatcher/utils.py	2012-12-15 00:15:28 +0000
@@ -163,13 +163,23 @@
 
 class logging_spawn(pexpect.spawn):
 
+    def __init__(self, command, timeout=30, logfile=None):
+        pexpect.spawn.__init__(
+            self, command, timeout=timeout, logfile=logfile)
+
+        # serial can be slow, races do funny things, so increase delay
+        self.delaybeforesend = 0.05
+
     def sendline(self, s=''):
-        logging.debug("sendline : %s" % s)
+        logging.debug("sendline : %s", s)
         return super(logging_spawn, self).sendline(s)
 
-    def send(self, *args, **kw):
-        logging.debug("send : %s" % args[0])
-        return super(logging_spawn, self).send(*args, **kw)
+    def send(self, string):
+        logging.debug("send : %s", string)
+        sent = 0
+        for char in string:
+            sent += super(logging_spawn, self).send(char)
+        return sent
 
     def expect(self, *args, **kw):
         # some expect should not be logged because it is so much noise.
@@ -183,9 +193,9 @@
             timeout = self.timeout
 
         if len(args) == 1:
-            logging.debug("expect (%d): '%s'" % (timeout, args[0]))
+            logging.debug("expect (%d): '%s'", timeout, args[0])
         else:
-            logging.debug("expect (%d): '%s'" % (timeout, str(args)))
+            logging.debug("expect (%d): '%s'", timeout, str(args))
 
         return super(logging_spawn, self).expect(*args, **kw)