Remove transaction when issues are created.

Remove transaction when issues are created.

This patch is a preparation for allocating ids for issues and patchsets programmatically. It's not possible to allocate ids within transactions.

[M-A Ruel, 2013-06-20 14:35:43]

lgtm

[techtonik, 2013-06-20 17:16:06]

Seems good to me. I don't know if it is worthy to do refactoring of engine API
right now considering that the change is necessary for the other purpose, but if
it is possible to find exceptions that engine uses, I'd prefer to have them
catched explicitly.

https://codereview.appspot.com/10434044/diff/1/codereview/views.py
File codereview/views.py (right):

https://codereview.appspot.com/10434044/diff/1/codereview/views.py#newcode1735
codereview/views.py:1735: except:
I can't say how bad I feel about the absence of user feedback in this code.
"Empty patchset != invalid patchset" and the latter can mean mistake in our own
tools. But considering that this logic was here before, this particular change
doesn't seem too bad to me.

As I am not ready to propose any alternative, do you think that inserting a TODO
item with a reference to http://code.google.com/p/rietveld/issues/detail?id=451
here will be better than nothing?

https://codereview.appspot.com/10434044/diff/1/codereview/views.py#newcode1738
codereview/views.py:1738: # happen there.
I prefer to have defined interface for interaction with engine instead of catch
all logic.

[Andi, 2013-06-20 18:28:31]

Committed as 1b7869f9a523

[Andi, 2013-06-20 18:30:42]

https://codereview.appspot.com/10434044/diff/1/codereview/views.py
File codereview/views.py (right):

https://codereview.appspot.com/10434044/diff/1/codereview/views.py#newcode1735
codereview/views.py:1735: except:
On 2013/06/20 17:16:07, techtonik wrote:
> I can't say how bad I feel about the absence of user feedback in this code.
> "Empty patchset != invalid patchset" and the latter can mean mistake in our
own
> tools. But considering that this logic was here before, this particular change
> doesn't seem too bad to me.
> 
> As I am not ready to propose any alternative, do you think that inserting a
TODO
> item with a reference to
http://code.google.com/p/rietveld/issues/detail?id=451
> here will be better than nothing?
> 

I don't like those catch all too. But we don't know right now what errors could
happen here. Neither ParsePatchSet nor SplitPatch have explicit exceptions.
However, ParsePatchSet was called without any exception handling for years now
when uploading new patchsets for an issue and I'm not aware of any exceptions.

[techtonik, 2013-06-20 21:53:38]

Ok. Let's assume this part is mature.