Questions after testtools merge

[Martin (gzlist)]

On 28/12/2009, Martin (gzlist) <gzlist at googlemail.com> wrote:
> Seeing if I can reproduce with a smaller subset currently, and not
> having any luck so far. I have confirmed it's a r4920 regression
> however.

Okay, I've got a minimal repo now:

C:\bzr\bzr\dev>bzr revert -r 4919 -q

C:\bzr\bzr\dev>python bzr selftest -s
bb.test_non_ascii.TestNonAscii.test_mkdir(iso-8859-2) -s
bb.test_version.TestVersionUnicodeOutput.test_unicode_bzr_home
testing: C:/bzr/bzr/dev/bzr
   C:\bzr\bzr\dev\bzrlib
   bzr-2.1.0dev5 python-2.4.4 Windows-XP-5.1.2600-SP3


----------------------------------------------------------------------
Ran 2 tests in 0.203s

OK
1 test skipped

C:\bzr\bzr\dev>bzr revert -r 4920 -q

C:\bzr\bzr\dev>python bzr selftest -s
bb.test_non_ascii.TestNonAscii.test_mkdir(iso-8859-2) -s
bb.test_version.TestVersionUnicodeOutput.test_unicode_bzr_home
bzr selftest: C:/bzr/bzr/dev/bzr
   C:\bzr\bzr\dev\bzrlib
   bzr-2.1.0dev5 python-2.4.4 Windows-XP-5.1.2600-SP3

[2/2 in 0s] blackbox.test_version.TestVersionUnicodeOutput.test_unicode_bzr_home
C:\bzr\bzr\dev>echo %ERRORLEVEL%
-1073741819

Haven't dug into this further yet, but I suspect that bazaar is
corrupting the environment block when doing the codepage-switching
attempts. Why testtools does this when the previous testing code
didn't I have no idea.

Martin