record_entry_contents needs tests for whether it actually records a new text

On Wed, 2009-08-26 at 05:26 +0000, Martin Pool wrote:
>
>
> >From talking to John about bug 415508: record_entry_contents has some
> tests that check it records data when it should but (I think) no tests
> that it doesn't store data when it shouldn't - so you can get the kind
> of misbehaviour reported by Frits.

it has tests over precisely when, and when not to store new texts.

The problem is that the interface lied to it.

-Rob

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Martin Pool wrote:
> Public bug reported:
>
>>From talking to John about bug 415508: record_entry_contents has some
> tests that check it records data when it should but (I think) no tests
> that it doesn't store data when it shouldn't - so you can get the kind
> of misbehaviour reported by Frits.

Well, even beyond 'record_entry_contents', commit in general doesn't
have any of these tests.

So while we are quite confident that things that should be there are
present (we would get bug reports, breakage, etc if that wasn't the
case). If we add too much, it just doesn't show up.

In Frits's case, this is exacerbated by the fact that "bzr log -v"
doesn't show these modifications, even though "bzr log -v FILE" *does*.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkqVREcACgkQJdeBCYSNAANUbgCg2fybcF0WMUvAj76PBpVkNzXX
vGgAoLZnbu0MCSdFF6sXMG3ZtrgQ4yXj
=KOnZ
-----END PGP SIGNATURE-----

On Wed, 2009-08-26 at 14:18 +0000, John A Meinel wrote:

> Well, even beyond 'record_entry_contents', commit in general doesn't
> have any of these tests.
>
> So while we are quite confident that things that should be there are
> present (we would get bug reports, breakage, etc if that wasn't the
> case). If we add too much, it just doesn't show up.

Huh? Check for unchanged in per_repository.test_commit_builder.

Hundreds of tests, complete set of permutations.

-Rob