On Fri, 2009-07-10 at 03:00 +0000, Aaron Bentley wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Robert Collins wrote:
> > The critical component was truely critical and has been cherrypicked
> > into lp production some time ago. The lp task would be appropriately
> > closed, I think.
>
> This is hitting us any time we use bundles to create/update branches.
> It's a pretty high-frequency, high-irritation event.
We should definitely solve your problem. However, I think thats a
different bug. ACF is a very generic error, its like 'segfault' for C
programs. Getting an ACF means something is wrong, but doesn't indicate
if its missing data or bad delta logic.
Bug 390563 - *this* bug - is 'when all required data is present, bzr
still fails to generate a stream'. And the fix that has been
cherrypicked should be robust except when the tree height changes.
We should debug the bundle based one. I think it is bug 393349, as JML
has suggested.
On Fri, 2009-07-10 at 03:00 +0000, Aaron Bentley wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Robert Collins wrote:
> > The critical component was truely critical and has been cherrypicked
> > into lp production some time ago. The lp task would be appropriately
> > closed, I think.
>
> This is hitting us any time we use bundles to create/update branches.
> It's a pretty high-frequency, high-irritation event.
We should definitely solve your problem. However, I think thats a
different bug. ACF is a very generic error, its like 'segfault' for C
programs. Getting an ACF means something is wrong, but doesn't indicate
if its missing data or bad delta logic.
Bug 390563 - *this* bug - is 'when all required data is present, bzr
still fails to generate a stream'. And the fix that has been
cherrypicked should be robust except when the tree height changes.
We should debug the bundle based one. I think it is bug 393349, as JML
has suggested.
-Rob