Update sqlite to 3.7.3.

Update sqlite to 3.7.3.

BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=68506

[James Hawkins, 2010-12-03 21:26:44]

ping

[Scott Hess - ex-Googler, 2010-12-03 23:21:44]

Make sure it passes the SQLite tests after our patches are applied.  I think
Linux is the easiest place to run them, there should be instructions in our
README.

Suggest that you annotate our README to indicate the status of the fts3 patches.
 Someone will probably have to go over them and compare fts3_write.c.  Sigh. 
Guess that's either Chris or me, probably.

AFAICT from the preprocessed/ files and parse.y, this doesn't introduce any new
syntax or settings which will cause us security issues.  Which is good :-).

[Adding mdm and cevans and dumi as random interested parties who should feel
free to say "And also run XXX!" at this point.]

http://codereview.chromium.org/5626002/diff/36007/chrome/browser/download/dow...
File chrome/browser/download/download_util_unittest.cc (right):

http://codereview.chromium.org/5626002/diff/36007/chrome/browser/download/dow...
chrome/browser/download/download_util_unittest.cc:441: EXPECT_TRUE(found)
Don't include this change with the overall change.

[James Hawkins, 2010-12-06 22:15:42]

On 2010/12/03 23:21:44, shess wrote:
> Make sure it passes the SQLite tests after our patches are applied.  I think
> Linux is the easiest place to run them, there should be instructions in our
> README.
> 

0 errors out of 74048 tests

> Suggest that you annotate our README to indicate the status of the fts3
patches.
>  Someone will probably have to go over them and compare fts3_write.c.  Sigh. 
> Guess that's either Chris or me, probably.
> 

Updated the README, and I'll add this note to the bug I'm opening for sec
review.

> AFAICT from the preprocessed/ files and parse.y, this doesn't introduce any
new
> syntax or settings which will cause us security issues.  Which is good :-).
> 
> [Adding mdm and cevans and dumi as random interested parties who should feel
> free to say "And also run XXX!" at this point.]
> 
>
http://codereview.chromium.org/5626002/diff/36007/chrome/browser/download/dow...
> File chrome/browser/download/download_util_unittest.cc (right):
> 
>
http://codereview.chromium.org/5626002/diff/36007/chrome/browser/download/dow...
> chrome/browser/download/download_util_unittest.cc:441: EXPECT_TRUE(found)
> Don't include this change with the overall change.

Removed in patch set 3 (uploading now).

[Scott Hess - ex-Googler, 2010-12-07 18:44:12]

I'm going with LGTM.  SQLite is pretty good about not pushing totally broken
stuff, and we branched 9.0 recently so this won't land on beta or stable anytime
soon.  I can keep an eye on their mailing list for reported brokenness.

This is to fix tests, right, not for new features?  So it won't be a problem
reverting if we find something subtle in a week or two?

[James Hawkins, 2010-12-07 19:24:02]

On 2010/12/07 18:44:12, shess wrote:
> I'm going with LGTM.  SQLite is pretty good about not pushing totally broken
> stuff, and we branched 9.0 recently so this won't land on beta or stable
anytime
> soon.  I can keep an eye on their mailing list for reported brokenness.
> 
> This is to fix tests, right, not for new features?  So it won't be a problem
> reverting if we find something subtle in a week or two?

Correct.

[Scott Hess - ex-Googler, 2010-12-07 21:07:34]

BTW - since lots of our SQLite usage is in the form of server threads being
peppered with IPC calls, SQLite tends to be an attractor for valgrind
suppressions.  I expect this will invalidate some of those.  Most definitely one
for watching the memory bots.