bzr selftest has many failures

[Martin Pool]

On 16 May 2005, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
> Martin Pool wrote:
> | On 15 May 2005, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
> |
> |>I ran testbzr, and it appeared internal tests passed.  But they hadn't.
> |
> |
> | Thanks, both the failures and the failure to report them are now fixed.
> 
> bzr selftest now reports:
> bzr: No module named tests

Ah, I had a .pyc left behind.  

This would be a great case for borrowing darcs' pre-checkin test
strategy:

  Make an export of the too-be-committed revision to a new temporary
  directory.  Run the test command there.  If it fails, abort the
  commit and show the error.

This traps several important cases compared to just testing in the
working directory:

 - depending on unknown/ignored files (as here)

 - partial commit that misses some dependencies

 - other bugs that let an incremental build succeed but not a clobber

> While I agree that blackbox testing is important, whitebox (unit)
> testing is also important.  It allows you to test the api with much
> smaller granularity than the blackbox approach.

I agree.  I removed these cases because they were at a large
granularity and so can be done just as well (or better) through the
command-line interface (and also they were rather outdated.)

-- 
Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20050517/bde895d4/attachment.pgp