On Wed, Aug 27, 2008 at 3:28 PM, Michael Nordman <[EMAIL PROTECTED]> wrote:
> ------------------------------------
> Line 319: // Hooray, it succeeded. Nothing more to do here.
> maybe invert the test, LOG("something went wrong") if anything doesn't work

On the flip side, it's nice that the current version uses parallel
logic in Step 1 and Step 2.  Feels like it could go either way.  I'm
leaning toward leaving as-is.

> ------------------------------------
> Line 337: File::Delete(full_path.c_str());  // ignore return value
> As files are really deleted, we may be able to actually delete the 
> corresponding
> row instead of leaving them marked as deleted. Not sure its worth it... just
> throwing this out there.

Yeah, not sure it's worth it.  I also don't know the other SQL well
enough to be sure it's 100% safe.  Again, leaning toward leaving
as-is.

> ========================================================================
> http://mondrian.corp.google.com/file/8116546///depot/googleclient/gears/opensource/gears/base/common/permissions_db.cc?a=1
> File 
> //depot/googleclient/gears/opensource/gears/base/common/permissions_db.cc 
> (snapshot 1)
> ------------------------------------
> Line 186: database_name_table_.DeleteOriginDatabases(origin);
> maybe rename this to DeleteDatabasesForOrigin for consistency with the 
> language
> used below

Done.

> ------------------------------------
> Line 192: }
> At this point, we may be able to delete data directory for this origin all
> together.
>
> There may be shortcut icons stored in here, otherwise Gears would have no 
> other
> data within this directory.
>
> Deleting the directory if its empty (or contains only empty directories) would
> be safe.

I think I'd have to write such a function, and test it well.  It
doesn't seem worthwhile, since this only affects empty directories.

Reply via email to