Only do the exclude-journal-from-Time-Machine-if-the-database-is-excluded thing for unix-flavored

Only do the exclude-journal-from-Time-Machine-if-the-database-is-excluded
thing for unix-flavored sqlite3 VFSes.

It's wrong to try to do this for the chromium_vfs sqlite3 VFS type. The
exclude code should only be used for databases that are actually on disk.
Under chromium_vfs, zFilename is a name but not a filesystem pathname, and
in the Chromium renderer process, direct filesystem access is forbidden.
The resulting CFURLRef objects did not have valid referents, and on Mac OS
X 10.7 ("Lion"), they resulted in use-after-free and double-free errors.

BUG=91068
TEST=With a clean profile, visit http://www.justgiving.nl/. The page should
     not sad tab. Nothing should be logged to the console. Previously,
     messages such as the following would be logged:

Google Chrome Helper(12345,0xabcdef00) malloc: *** error for object 0x4545450: incorrect checksum for freed object - object was probably modified after being freed.
Google Chrome Helper(12345,0xabcdef00) malloc: *** error for object 0x4545450: double free
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=95435

[mrossetti, 2011-08-04 16:34:49]

LGTM

[mrossetti, 2011-08-04 16:37:23]

Nits.

http://codereview.chromium.org/7511011/diff/1/third_party/sqlite/amalgamation...
File third_party/sqlite/amalgamation/sqlite3.c (right):

http://codereview.chromium.org/7511011/diff/1/third_party/sqlite/amalgamation...
third_party/sqlite/amalgamation/sqlite3.c:41643: +         &&
memcmp(pVfs->zName, "unix", 4) == 0
Nit: No spaces around '==' here and two places in next line.

http://codereview.chromium.org/7511011/diff/1/third_party/sqlite/mac_time_mac...
File third_party/sqlite/mac_time_machine.patch (right):

http://codereview.chromium.org/7511011/diff/1/third_party/sqlite/mac_time_mac...
third_party/sqlite/mac_time_machine.patch:58: +         && memcmp(pVfs->zName,
"unix", 4) == 0
Nit: No spaces around '==' here and two places in next line.

http://codereview.chromium.org/7511011/diff/1/third_party/sqlite/src/src/pager.c
File third_party/sqlite/src/src/pager.c (right):

http://codereview.chromium.org/7511011/diff/1/third_party/sqlite/src/src/page...
third_party/sqlite/src/src/pager.c:5212: && memcmp(pVfs->zName, "unix", 4) == 0
Nit: No spaces around '==' here and two places in next line.

[Mark Mentovai, 2011-08-04 16:43:09]

Nits picked.

[Scott Hess - ex-Googler, 2011-08-04 18:43:28]

LGTM.

Debugging this was either inspired, or you're just insane, both are good, but I
wonder: Does this mean that WebDatabase journal files will get time-machined? 
It sounds like the code wasn't previously working, so no regression, I'm just
wondering if I should log a new bug.

[Mark Mentovai, 2011-08-04 19:25:08]

Inspirational insanity.

WebDatabase journal files will get TimeMachined because their main databases are
TimeMachined. We don’t ever exclude a -journal independently, we only exclude
them when the main database is excluded. Nobody’s ever suggested that we exclude
WebDatabase databases. I don’t think we should exclude them. Do you?

(If we need to go back and forth on this for more than another round or two, I
agree we should discuss it on a new bug.)

[Scott Hess - ex-Googler, 2011-08-04 19:36:21]

Hmm.  OK, that makes sense.  I don't think this problem is nearly as
keen for WebDatabase files as it is for things like history.  There is
probably some justification for not time-machining ANY web data files,
but I don't know if it even warrants discussion.

-scott


On Thu, Aug 4, 2011 at 12:25 PM,  <mark@chromium.org> wrote:
> Inspirational insanity.
>
> WebDatabase journal files will get TimeMachined because their main databases
> are
> TimeMachined. We don’t ever exclude a -journal independently, we only
> exclude
> them when the main database is excluded. Nobody’s ever suggested that we
> exclude
> WebDatabase databases. I don’t think we should exclude them. Do you?
>
> (If we need to go back and forth on this for more than another round or two,
> I
> agree we should discuss it on a new bug.)
>
> http://codereview.chromium.org/7511011/
>