Android handler: Fix name collision.

Android handler: Fix name collision.

The android handler was using the android framework to create temporary
files. This creates predictible names and ends up creating 2 libraries
with the same name. This is then an issue with ldopen that will not open
those independently.

This CL changes the implementation to create names that will not be
predictible.

R=blundell@chromium.org

Committed: https://chromium.googlesource.com/external/mojo/+/010222933db7567616e6f3d4c26a2439e1836664

[blundell, 2015-05-13 09:16:43]

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
File services/url_response_disk_cache/url_response_disk_cache_impl.cc (left):

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
services/url_response_disk_cache/url_response_disk_cache_impl.cc:324:
WriteFile(GetExtractedSentinel(dir), nullptr, 0);
Am I understanding correctly that this code was previously treating the consumer
cache directory as the base directory? It's surprising that the tests were
passing if so.

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
File services/url_response_disk_cache/url_response_disk_cache_impl.cc (right):

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
services/url_response_disk_cache/url_response_disk_cache_impl.cc:51: if
(!std::remove_pointer<DataType>::type::Validate(&input[0],
What exactly is this doing?

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
services/url_response_disk_cache/url_response_disk_cache_impl.cc:187: //
Obsolete entry are invalidated.
nit: s/entry/entries

https://codereview.chromium.org/1133283002/diff/20001/shell/android/android_h...
File shell/android/android_handler.cc (right):

https://codereview.chromium.org/1133283002/diff/20001/shell/android/android_h...
shell/android/android_handler.cc:175: int fd =
open(temporary_file.value().c_str(), O_CREAT | O_EXCL, 0600);
Are you doing this construction in C++ rather than Java so that you can do this
atomic check that the file doesn't already exist?

[qsr, 2015-05-13 09:32:05]

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
File services/url_response_disk_cache/url_response_disk_cache_impl.cc (left):

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
services/url_response_disk_cache/url_response_disk_cache_impl.cc:324:
WriteFile(GetExtractedSentinel(dir), nullptr, 0);
On 2015/05/13 09:16:43, blundell wrote:
> Am I understanding correctly that this code was previously treating the
consumer
> cache directory as the base directory? It's surprising that the tests were
> passing if so.

It was only doing so when sending the cache_dir, so the cache_dir was
{base_path}/consumer_cache/consumer_cache

This is working, but ugly.

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
File services/url_response_disk_cache/url_response_disk_cache_impl.cc (right):

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
services/url_response_disk_cache/url_response_disk_cache_impl.cc:51: if
(!std::remove_pointer<DataType>::type::Validate(&input[0],
On 2015/05/13 09:16:43, blundell wrote:
> What exactly is this doing?

This is using the generated validate method to check that the byte we get from
the file are a valid entry.

https://codereview.chromium.org/1133283002/diff/20001/services/url_response_d...
services/url_response_disk_cache/url_response_disk_cache_impl.cc:187: //
Obsolete entry are invalidated.
On 2015/05/13 09:16:43, blundell wrote:
> nit: s/entry/entries

Done.

https://codereview.chromium.org/1133283002/diff/20001/shell/android/android_h...
File shell/android/android_handler.cc (right):

https://codereview.chromium.org/1133283002/diff/20001/shell/android/android_h...
shell/android/android_handler.cc:175: int fd =
open(temporary_file.value().c_str(), O_CREAT | O_EXCL, 0600);
On 2015/05/13 09:16:43, blundell wrote:
> Are you doing this construction in C++ rather than Java so that you can do
this
> atomic check that the file doesn't already exist?

Nope, I do this in C++ because:
1) I now do not trust the random generator
2) There is no easy way to convert bytes to hexadecimal values

[blundell, 2015-05-13 09:38:01]

lgtm

[qsr, 2015-05-13 10:27:38]

Committed patchset #4 (id:60001) manually as
010222933db7567616e6f3d4c26a2439e1836664 (presubmit successful).