why does bazaar not add a file automatically on the first commit

[John Arbash Meinel]

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

Wichmann, Mats D wrote:
|> (A more appropriate fix for this sort of thing is to have your test
|> suite passing on mainline... just a thought)
|
| It's not a bad thought; something could be doing sanity checks
| during the upstream merge (we use pqm, but it doesn't do checks).
| Just that would slow down a cumbersome process even more but
| it doesn't make the idea wrong....
|
| We do have automated tests, they run on the "head" branch
| but this was a case on an older (maintenace) branch so it
| got missed.
|
| Still, I prefer explicit to implicit on file removals.
| The merge with any associated mandatory tests might happen
| much later when the context is not so clear, I'd like
| complaints when the local commit happens.
|
|

I'm honestly with you there, just explaining what I consider the "real" approach
to working-tree correctness.

My particular reason for not wanting auto-remove is renames:

bzr add foo
bzr commit -m 'foo'
mv foo bar

bzr commit -m "silently remove foo, and ignore bar"


I work around this with:

[ALIASES]
commit = commit --strict

So at least unknown files will prevent a commit unless I force it with
(--no-strict).

It is a little odd that partial commits that don't touch those files also fail,
but I'm willing to live with that. (I mostly do it for when I create test cases,
etc. and forget to add them before I commit.)

John
=:->

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

iEYEARECAAYFAkgOpuIACgkQJdeBCYSNAAMwxgCghk+5FgupY8SA/Zfv86ifIqLT
gyMAoNmcvkHudLley06r8SuyNJ6YyqIv
=qeLz
-----END PGP SIGNATURE-----