Enable binary file upload.

Enable binary file upload.

It seems to just work.

R=dpranke@chromium.org
BUG=23608
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=222431

[Dirk Pranke, 2013-09-11 00:46:32]

lgtm. let's give it a shot!

[commit-bot: I haz the power, 2013-09-11 00:46:51]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/chrisphan@chromium.org/23766025/4001

[commit-bot: I haz the power, 2013-09-11 00:49:18]

Change committed as 222431

[jam, 2013-09-11 01:51:29]

Might be worth getting Marc-Antoine's input since he disabled this in
https://codereview.chromium.org/11016032/

[Dirk Pranke, 2013-09-11 04:14:58]

On 2013/09/11 01:51:29, jam wrote:
> Might be worth getting Marc-Antoine's input since he disabled this in
> https://codereview.chromium.org/11016032/

Sure, that's a good suggestion.

That said, I do recall discussing this issue w/ M-A and Robbie a few times over
the past year, and the general consensus was that we thought at this point
things should more-or-less work, and someone just needed to do the testing.
Chris did the testing, and things seem to work for at least some of the common
cases (as evidenced by my first try job after this landed working correctly).

It's also possible that we have fewer configurations to either test or at least
worry about supporting now that we're mostly git-based (e.g., neither git-try
nor gcl try are supported for Blink).

Mostly I landed this change so we could soft-launch this to do more realistic
testing over the next day or two. I won't be surprised if we want to revert it.

[M-A Ruel, 2013-09-11 13:54:51]

The main issue I had seen at the time was silent data corruption, that's why I
was highly concerned. From memory, it had shown up as a file cut of by 4 bytes
or empty file. Stay alert with this kind of issue, especially test both cases of
Rietveld decides the binary file is or is not an image, as there's different
internal handling, which is separate from upload.py image detection.

[Dirk Pranke, 2013-09-11 16:06:03]

On 2013/09/11 13:54:51, M-A Ruel wrote:
> The main issue I had seen at the time was silent data corruption, that's why I
> was highly concerned. 

Thanks; we should definitely be on the lookout for this. 

The good news is that at least in the blink case, most binaries are PNGs and
we'll catch problems in the layout_tests (since we'll be verifying the PNGs
match what we are generating). We may not want to encourage people to use this
for arbitrary commits of binary files just yet.