AU dev-server: Add a sha256 hash to the updatecheck response.

AU dev-server: Add a sha256 hash to the updatecheck response.

BUG=6580
TEST=tested with an update_engine that takes SHA-256.

Change-Id: Id7d507fe65f8eb32d77b14b76b80d3d7ba52182b

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=91436cb

[adlr, 2010-09-23 20:12:47]

LGTM w/ random idea...

http://codereview.chromium.org/3455018/diff/1/2
File autoupdate.py (right):

http://codereview.chromium.org/3455018/diff/1/2#newcode135
autoupdate.py:135: def _GetSHA256(self, update_path):
if you want to be fancy, and use bash, you can read the file once and send it to
both copies of openssl. less disk i/o and takes advantage of dual-core, but may
not really mater...


#!/bin/bash

coproc (openssl dgst -sha256 -binary | openssl base64 > /tmp/sha256)
cat $1 | tee /proc/${COPROC_PID}/fd/0 | openssl sha1 -binary | openssl base64 >
/tmp/sha1
eval "exec ${COPROC[1]}>&-"
wait ${COPROC_PID}
echo SHA265: $(cat /tmp/sha256)
echo SHA1: $(cat /tmp/sha1)

[petkov, 2010-09-23 20:31:08]

On 2010/09/23 20:12:47, adlr wrote:
> LGTM w/ random idea...
> 
> http://codereview.chromium.org/3455018/diff/1/2
> File autoupdate.py (right):
> 
> http://codereview.chromium.org/3455018/diff/1/2#newcode135
> autoupdate.py:135: def _GetSHA256(self, update_path):
> if you want to be fancy, and use bash, you can read the file once and send it
to
> both copies of openssl. less disk i/o and takes advantage of dual-core, but
may
> not really mater...
> 
> 
> #!/bin/bash
> 
> coproc (openssl dgst -sha256 -binary | openssl base64 > /tmp/sha256)
> cat $1 | tee /proc/${COPROC_PID}/fd/0 | openssl sha1 -binary | openssl base64
>
> /tmp/sha1
> eval "exec ${COPROC[1]}>&-"
> wait ${COPROC_PID}
> echo SHA265: $(cat /tmp/sha256)
> echo SHA1: $(cat /tmp/sha1)

Good idea although I'm not sure the extra complexity is worth it given that this
runs on the server. Also, I think this is equivalent to:

cat test | tee > /dev/null \
       >( openssl dgst -sha256 -binary | openssl base64 > /tmp/sha256 ) \
       >( openssl sha1 -binary | openssl base64 > /tmp/sha1 )
echo SHA265: $(cat /tmp/sha256)
echo SHA1: $(cat /tmp/sha1)

[adlr, 2010-09-23 21:17:55]

oh, that works! much nicer

On Thu, Sep 23, 2010 at 1:31 PM, <petkov@chromium.org> wrote:

> On 2010/09/23 20:12:47, adlr wrote:
>
>> LGTM w/ random idea...
>>
>
>  http://codereview.chromium.org/3455018/diff/1/2
>> File autoupdate.py (right):
>>
>
>  http://codereview.chromium.org/3455018/diff/1/2#newcode135
>> autoupdate.py:135: def _GetSHA256(self, update_path):
>> if you want to be fancy, and use bash, you can read the file once and send
>> it
>>
> to
>
>> both copies of openssl. less disk i/o and takes advantage of dual-core,
>> but
>>
> may
>
>> not really mater...
>>
>
>
>  #!/bin/bash
>>
>
>  coproc (openssl dgst -sha256 -binary | openssl base64 > /tmp/sha256)
>> cat $1 | tee /proc/${COPROC_PID}/fd/0 | openssl sha1 -binary | openssl
>> base64
>>
>
>  /tmp/sha1
>> eval "exec ${COPROC[1]}>&-"
>> wait ${COPROC_PID}
>> echo SHA265: $(cat /tmp/sha256)
>> echo SHA1: $(cat /tmp/sha1)
>>
>
> Good idea although I'm not sure the extra complexity is worth it given that
> this
> runs on the server. Also, I think this is equivalent to:
>
> cat test | tee > /dev/null \
>       >( openssl dgst -sha256 -binary | openssl base64 > /tmp/sha256 ) \
>       >( openssl sha1 -binary | openssl base64 > /tmp/sha1 )
>
> echo SHA265: $(cat /tmp/sha256)
> echo SHA1: $(cat /tmp/sha1)
>
>
> http://codereview.chromium.org/3455018/show
>

[Nick Sanders, 2010-09-24 00:52:09]

http://codereview.chromium.org/3455018/diff/1/2
File autoupdate.py (right):

http://codereview.chromium.org/3455018/diff/1/2#newcode141
autoupdate.py:141: def GetUpdatePayload(self, hash, sha256, size, url):
Can you test this with factory install?

http://codereview.chromium.org/3455018/diff/1/2#newcode421
autoupdate.py:421: return self.GetUpdatePayload(checksum, size, url)
Missing arg

[petkov, 2010-09-24 16:03:14]

Nick, I'll submit soon but it would be great if you could test the factory flow.
I don't expect any issues but you never know.

http://codereview.chromium.org/3455018/diff/1/2
File autoupdate.py (right):

http://codereview.chromium.org/3455018/diff/1/2#newcode135
autoupdate.py:135: def _GetSHA256(self, update_path):
On 2010/09/23 20:12:47, adlr wrote:
> if you want to be fancy, and use bash, you can read the file once and send it
to
> both copies of openssl. less disk i/o and takes advantage of dual-core, but
may
> not really mater...
> 
> 
> #!/bin/bash
> 
> coproc (openssl dgst -sha256 -binary | openssl base64 > /tmp/sha256)
> cat $1 | tee /proc/${COPROC_PID}/fd/0 | openssl sha1 -binary | openssl base64
>
> /tmp/sha1
> eval "exec ${COPROC[1]}>&-"
> wait ${COPROC_PID}
> echo SHA265: $(cat /tmp/sha256)
> echo SHA1: $(cat /tmp/sha1)
> 

Added a TODO. This runs on the server so it's not worth spending too much time
optimizing unless it becomes an issue.

http://codereview.chromium.org/3455018/diff/1/2#newcode141
autoupdate.py:141: def GetUpdatePayload(self, hash, sha256, size, url):
On 2010/09/24 00:52:10, Nick Sanders wrote:
> Can you test this with factory install?

Don't know how. Could you please test it?

http://codereview.chromium.org/3455018/diff/1/2#newcode421
autoupdate.py:421: return self.GetUpdatePayload(checksum, size, url)
On 2010/09/24 00:52:10, Nick Sanders wrote:
> Missing arg

Done.

[Nick Sanders, 2010-09-25 00:34:10]

I'm in China now, so unfortunately I can't easily test this.

Factory install is the use of this software that's most critical to keep
working, so please don't check in code without running it first. I've sent you
email off-review on how to do this.

On 2010/09/24 16:03:14, petkov wrote:
> Nick, I'll submit soon but it would be great if you could test the factory
flow.
> I don't expect any issues but you never know.
> 
> http://codereview.chromium.org/3455018/diff/1/2
> File autoupdate.py (right):
> 
> http://codereview.chromium.org/3455018/diff/1/2#newcode135
> autoupdate.py:135: def _GetSHA256(self, update_path):
> On 2010/09/23 20:12:47, adlr wrote:
> > if you want to be fancy, and use bash, you can read the file once and send
it
> to
> > both copies of openssl. less disk i/o and takes advantage of dual-core, but
> may
> > not really mater...
> > 
> > 
> > #!/bin/bash
> > 
> > coproc (openssl dgst -sha256 -binary | openssl base64 > /tmp/sha256)
> > cat $1 | tee /proc/${COPROC_PID}/fd/0 | openssl sha1 -binary | openssl
base64
> >
> > /tmp/sha1
> > eval "exec ${COPROC[1]}>&-"
> > wait ${COPROC_PID}
> > echo SHA265: $(cat /tmp/sha256)
> > echo SHA1: $(cat /tmp/sha1)
> > 
> 
> Added a TODO. This runs on the server so it's not worth spending too much time
> optimizing unless it becomes an issue.
> 
> http://codereview.chromium.org/3455018/diff/1/2#newcode141
> autoupdate.py:141: def GetUpdatePayload(self, hash, sha256, size, url):
> On 2010/09/24 00:52:10, Nick Sanders wrote:
> > Can you test this with factory install?
> 
> Don't know how. Could you please test it?
> 
> http://codereview.chromium.org/3455018/diff/1/2#newcode421
> autoupdate.py:421: return self.GetUpdatePayload(checksum, size, url)
> On 2010/09/24 00:52:10, Nick Sanders wrote:
> > Missing arg
> 
> Done.

[petkov, 2010-09-27 19:13:43]

+cc ericli,seano: you'll need to propagate this patch to the test server's AU
server otherwise update_engine-based image update will stop working once
http://codereview.chromium.org/3419018/show is in.