Fix a problem in fts3_write.c causing stack memory to be referenced after it is out of scope

Upstream commit in trunk:
http://www.sqlite.org/src/info/f9c4a7c8f4

Upstream commits in branch-3.7.6:
http://www.sqlite.org/src/info/a61786e760 (third hunk)
http://www.sqlite.org/src/info/e6a04bf9f1 (correction)

I have left out the comment addition seen in the trunk commit, because the backport to the 3.7.6 branch doesn't include it.

[Patch from Evangelos Foutras <evangelos@foutrelis.com>; ]

BUG=122525
TEST=distros using gcc4.7 shouldn't crash.  see bug.
R=shess@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=141687

[Paweł Hajdan Jr., 2012-05-09 06:02:44]

Scott, could you take a look? This fixes a gcc-4.7 crash issue.

[Scott Hess - ex-Googler, 2012-05-18 21:24:15]

On 2012/05/09 06:02:44, Paweł Hajdan Jr. wrote:
> Scott, could you take a look? This fixes a gcc-4.7 crash issue.

LGTM.

I'm surprised some automated tool hasn't called out the problem earlier!

In verifying that the patch applied cleanly, I noticed that my amalgamation
files contained a hot patch we had done to prove out a shutdown problem.  I had
posted a CL to revert it, but forgotten to check it in, so I just clicked commit
on that.  So when committing this, double-check that
http://codereview.chromium.org/9845034/ didn't sneak back in or something.  I
don't think it will, but I think big files like sqlite3.c are sometimes handled
differently by the system.

[Scott Hess - ex-Googler, 2012-06-08 19:48:53]

I pressed the commit-queue on this, but it didn't go in.  I'm going to
land it manually.

On Fri, May 18, 2012 at 2:24 PM,  <shess@chromium.org> wrote:
> On 2012/05/09 06:02:44, Paweł Hajdan Jr. wrote:
>>
>> Scott, could you take a look? This fixes a gcc-4.7 crash issue.
>
>
> LGTM.
>
> I'm surprised some automated tool hasn't called out the problem earlier!
>
> In verifying that the patch applied cleanly, I noticed that my amalgamation
> files contained a hot patch we had done to prove out a shutdown problem.  I
> had
> posted a CL to revert it, but forgotten to check it in, so I just clicked
> commit
> on that.  So when committing this, double-check that
> http://codereview.chromium.org/9845034/ didn't sneak back in or something.
>  I
> don't think it will, but I think big files like sqlite3.c are sometimes
> handled
> differently by the system.
>
> http://codereview.chromium.org/10387026/

[Scott Hess - ex-Googler, 2012-06-08 19:57:50]

Aha.  I bet we don't have a contributor thingie on-file for you.

So, one question is: Do you want to be a contributor.  Info at the top of
http://dev.chromium.org/developers/contributing-code

If not, let me know.  I can just pull the patch from SQLite myself.

Thanks,
scott


On 2012/06/08 19:48:53, shess wrote:
> I pressed the commit-queue on this, but it didn't go in.  I'm going to
> land it manually.
> 
> On Fri, May 18, 2012 at 2:24 PM,  <mailto:shess@chromium.org> wrote:
> > On 2012/05/09 06:02:44, Paweł Hajdan Jr. wrote:
> >>
> >> Scott, could you take a look? This fixes a gcc-4.7 crash issue.
> >
> >
> > LGTM.
> >
> > I'm surprised some automated tool hasn't called out the problem earlier!
> >
> > In verifying that the patch applied cleanly, I noticed that my amalgamation
> > files contained a hot patch we had done to prove out a shutdown problem.
&nbsp;I
> > had
> > posted a CL to revert it, but forgotten to check it in, so I just clicked
> > commit
> > on that. &nbsp;So when committing this, double-check that
> > http://codereview.chromium.org/9845034/ didn't sneak back in or something.
> > &nbsp;I
> > don't think it will, but I think big files like sqlite3.c are sometimes
> > handled
> > differently by the system.
> >
> > http://codereview.chromium.org/10387026/

[Evangelos Foutras, 2012-06-08 20:15:55]

On 2012/06/08 19:57:50, shess wrote:
> Aha.  I bet we don't have a contributor thingie on-file for you.

True, I haven't contributed code before.

> So, one question is: Do you want to be a contributor.  Info at the top of
> http://dev.chromium.org/developers/contributing-code

Sure, I'm willing to follow the proper procedure.

I just signed [1] (not sure how quickly that gets processed). Should I also add
my information to the AUTHORS file and submit a new patch set?

Thanks.

[1] http://code.google.com/legal/individual-cla-v1.0.html

[Evangelos Foutras, 2012-06-08 21:10:33]

On 2012/06/08 20:15:55, foutrelis wrote:
> On 2012/06/08 19:57:50, shess wrote:
> > Aha.  I bet we don't have a contributor thingie on-file for you.
> 
> True, I haven't contributed code before.
> 
> > So, one question is: Do you want to be a contributor.  Info at the top of
> > http://dev.chromium.org/developers/contributing-code
> 
> Sure, I'm willing to follow the proper procedure.
> 
> I just signed [1] (not sure how quickly that gets processed). Should I also
add
> my information to the AUTHORS file and submit a new patch set?
> 
> Thanks.
> 
> [1] http://code.google.com/legal/individual-cla-v1.0.html

I have tried following the instructions on [1] as closely as I can. The second
patch set includes the change to AUTHORS.

Let me know if I missed something!

[1] http://dev.chromium.org/developers/contributing-code

[Scott Hess - ex-Googler, 2012-06-11 17:00:24]

On 2012/06/08 21:10:33, foutrelis wrote:
> On 2012/06/08 20:15:55, foutrelis wrote:
> > On 2012/06/08 19:57:50, shess wrote:
> > > Aha.  I bet we don't have a contributor thingie on-file for you.
> > 
> > True, I haven't contributed code before.
> > 
> > > So, one question is: Do you want to be a contributor.  Info at the top of
> > > http://dev.chromium.org/developers/contributing-code
> > 
> > Sure, I'm willing to follow the proper procedure.
> > 
> > I just signed [1] (not sure how quickly that gets processed). Should I also
> add
> > my information to the AUTHORS file and submit a new patch set?
> > 
> > Thanks.
> > 
> > [1] http://code.google.com/legal/individual-cla-v1.0.html
> 
> I have tried following the instructions on [1] as closely as I can. The second
> patch set includes the change to AUTHORS.
> 
> Let me know if I missed something!
> 
> [1] http://dev.chromium.org/developers/contributing-code

OK, I'm doing the try run described at the contributing-code page.  I'm not sure
how ancient that description is, maybe I'm supposed to use commit-queue, but
it's been a month, I doubt waiting until this afternoon will materially impact
world peace.

[Scott Hess - ex-Googler, 2012-06-12 18:09:54]

On 2012/06/11 17:00:24, shess wrote:
> OK, I'm doing the try run described at the contributing-code page.  I'm not
sure
> how ancient that description is, maybe I'm supposed to use commit-queue, but
> it's been a month, I doubt waiting until this afternoon will materially impact
> world peace.

For sheriffs who arrive here after I break things: I ran trybots yesterday, they
completed fine.  For some reason, the tries were not associated with the CL?  I
don't know why not.  I pressed the commit button, and it again hasn't sent me
email or anything, and the button has stayed set.  So I'm landing it manually.